Marjorie Heins, Christina Cho and Ariel Feldman Internet Filters A P u b l i c P o l i c y R e p o r t Second edition, fully revised and updated with a new introduction The Brennan Center for Justice, founded in 1995, unites thinkers and advocates in pursuit of a vision of inclusive and effective democracy. The Free Expression Policy Project founded in 2000, provides research and advocacy on free speech, copyright, and media democracy issues. FEPP joined the Brennan Center in 2004. Michael Waldman Executive Director Deborah Goldberg Director Democracy Program Marjorie Heins Coordinator Free Expression Policy Project The Brennan Center is grateful to the Robert Sterling Clark Foundation, the Nathan Cummings Foundation, the Rockefeller Foundation, and the Andy Warhol Foundation for the Visual Arts for support of the Free Expression Policy Project. Thanks to Kristin Glover, Judith Miller, Neema Trivedi, Samantha Frederickson, Jon Blitzer, and Rachel Nusbaum for research assistance. 2006. This work is covered by a Creative Commons “Attribution – No Derivatives – Noncommercial” License. It may be reproduced in its entirety as long as the Brennan Center for Justice, Free Expression Policy Project is credited, a link to the Project’s Web site is provided, and no charge is imposed. The report may not be reproduced in part or in altered form, or if a fee is charged, without our permission (except, of course, for “fair use”). Please let us know if you reprint. Cover illustration: © 2006 Lonni Sue Johnson Contents Executive Summary • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • i Introduction To The Second Edition The Origins of Internet Filtering • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 The “Children’s Internet Protection Act” (CIPA) • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 2 Living with CIPA • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 4 Filtering Studies During and After 2001• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 7 The Continuing Challenge • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 8 I. The 2001 Research Scan Updated: Over- And Underblocking By Internet Filters America Online Parental Controls • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 9 Bess • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 10 ClickSafe • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 14 Cyber Patrol • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 14 Cyber Sentinel • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 21 CYBERsitter • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 22 FamilyClick • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 25 I-Gear • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 26 Internet Guard Dog • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 28 Net Nanny • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 29 Net Shepherd • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 30 Norton Internet Security • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 31 SafeServer • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 31 SafeSurf • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 32 SmartFilter • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 32 SurfWatch • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 35 We-Blocker • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 38 WebSENSE • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 38 X-Stop • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 39 II. Research During and After 2001 Introduction: The Resnick Critique • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 45 Report for the Australian Broadcasting Authority • • • • • • • • • • • • • • • • • • • • • • • • • • • • 46 “Bess Won’t Go There” • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 49 Report for the European Commission: Currently Available COTS Filtering Tools • • • 50 Report for the European Commission: Filtering Techniques and Approaches • • • • • • • 52 Reports From the CIPA Litigation • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 53 Two Reports by Peacefire More Sites Blocked by Cyber Patrol • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 60 WebSENSE Examined • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 61 Two Reports by Seth Finkelstein BESS vs. Image Search Engines • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 61 BESS’s Secret Loophole • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 61 The Kaiser Family Foundation: Blocking of Health Information • • • • • • • • • • • • • • • • • 62 Two Studies From the Berkman Center for Internet and Society Web Sites Sharing IP Addresses • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 64 Empirical Analysis of Google SafeSearch • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 65 Electronic Frontier Foundation/Online Policy Group Study • • • • • • • • • • • • • • • • • • • • 66 American Rifleman • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 67 Colorado State Library • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 68 OpenNet Initiative • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 68 Rhode Island ACLU • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 69 Consumer Reports • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 69 Lynn Sutton PhD Dissertation: Experiences of High School Students Conducting Term Paper Research • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 70 Computing Which? Magazine • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 71 PamRotella.com: Experiences With iPrism • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 71 New York Times: SmartFilter Blocks Boing Boing • • • • • • • • • • • • • • • • • • • • • • • • • • • • 72 Conclusion and Recommendations • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 73 Bibliography • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 74 Executive Summary Every new technology brings with it both “tasteless/gross,” or “lifestyle”—their products excitement and anxiety. No sooner was the In- arbitrarily and irrationally blocked many Web ternet upon us in the 1990s than anxiety arose pages that had no relation to the disapproved over the ease of accessing pornography and content categories. For example: other controversial content. In response, en- • Net Nanny, SurfWatch, CYBERsitter, and trepreneurs soon developed filtering products. Bess blocked House Majority Leader Rich- By the end of the decade, a new industry had ard “Dick” Armey’s official Web site upon emerged to create and market Internet filters. detecting the word “dick.” These filters were highly imprecise. The • SmartFilter blocked the Declaration of problem was intrinsic to filtering technology. Independence, Shakespeare’s complete The sheer size of the Internet meant that iden- plays, Moby Dick, and Marijuana: Facts for tifying potentially offensive content had to be Teens, a brochure published by the National done mechanically, by matching “key” words Institute on Drug Abuse. and phrases; hence, the blocking of Web sites for “Middlesex County,” “Beaver College,” • SurfWatch blocked the human rights and “breast cancer”—just three of the bet- site Algeria Watch and the University of ter-known among thousands of examples of Kansas’s Archie R. Dykes Medical Library overly broad filtering. Internet filters were (upon detecting the word “dykes”). crude and error-prone because they catego- • CYBERsitter blocked a news item on the rized expression without regard to its context, Amnesty International site after detecting meaning, and value. the phrase “least 21.” (The offending sen- Some policymakers argued that these inac- tence described “at least 21” people killed curacies were an acceptable cost of keeping or wounded in Indonesia.) the Internet safe, especially for kids. Oth- • X-Stop blocked Carnegie Mellon Universi- ers—including many librarians, educators, and ty’s Banned Books page, the “Let’s Have an civil libertarians—argued that the cost was Affair” catering company, and, through its too high. To help inform this policy debate, “foul word” function, searches for Bastard the Free Expression Policy Project (FEPP) Out of Carolina and “The Owl and the published a report in the fall of 2001 sum- Pussy Cat.” marizing the results of more than 70 empirical studies on the performance of Internet filters. Despite such consistently irrational results, These studies ranged from anecdotal accounts the Internet filtering business continued to of blocked sites to extensive research applying grow. Schools and offices installed filters on social-science methods. their computers, and public libraries came under pressure to do so. In December 2000, Nearly every study revealed substantial over- President Bill Clinton signed the “Children’s blocking. That is, even taking into account Internet Protection Act,” mandating filters in that filter manufacturers use broad and vague all schools and libraries that receive federal aid blocking categories—for example, “violence,” for Internet connections. The Supreme Court Brennan Center for Justice upheld this law in 2003 despite extensive • WebSENSE blocked “Keep Nacogdoches evidence that filtering products block tens of Beautiful,” a Texas cleanup project, under thousands of valuable, inoffensive Web pages. the category of “sex,” and The Shoah Proj- ect, a Holocaust remembrance page, under The widespread use of filters the category of “racism/hate.” presents a serious threat to • Bess blocked all Google and AltaVista im- age searches as “pornography.” our most fundamental free • Google’s SafeSearch blocked congress.gov expression values. and shuttle.nasa.gov; a chemistry class at Middlebury College; Vietnam War materi- In 2004, FEPP, now part of the Brennan als at U.C.-Berkeley; and news articles from Center for Justice at N.Y.U. School of Law, the New York Times and Washington Post. decided to update the Internet Filters report—a project that continued through early 2006. The conclusion of the revised and updated We found several large studies published dur- Internet Filters: A Public Policy Report is that ing or after 2001, in addition to new, smaller- the widespread use of filters presents a serious scale tests of filtering products. Studies by the threat to our most fundamental free expres- U.S. Department of Justice, the Kaiser Family sion values. There are much more effective Foundation, and others found that despite ways to address concerns about offensive improved technology and effectiveness in Internet content. Filters provide a false sense blocking some pornographic content, filters of security, while blocking large amounts of are still seriously flawed. They continue to important information in an often irrational deprive their users of many thousands of valu- or biased way. Although some may say that able Web pages, on subjects ranging from war the debate is over and that filters are now a and genocide to safer sex and public health. fact of life, it is never too late to rethink bad Among the hundreds of examples: policy choices. ii Internet Filters: A Public Policy Report Introduction to the Second Edition The Origins of Internet Filtering rating” by filter manufacturers. Because of The Internet has transformed human commu- the Internet’s explosive growth (by 2001, nication. World Wide Web sites on every con- more than a billion Web sites, many of them ceivable topic, e-newsletters and listservs, and changing daily), and the consequent in- billions of emails racing around the planet ability of filtering company employees to daily have given us a wealth of information, evaluate even a tiny fraction of it, third-party ideas, and opportunities for communication rating had to rely on mechanical blocking never before imagined. As the U.S. Supreme by key words or phrases such as “over 18,” Court put it in 1997, “the content on the “breast,” or “sex.” The results were not dif- Internet is as diverse as human thought.” ficult to predict: large quantities of valuable information and literature, particularly about Not all of this online content is accurate, health, sexuality, women’s rights, gay and pleasant, or inoffensive. Virtually since the lesbian issues, and other important subjects, arrival of the Internet, concerns have arisen were blocked. about minors’ access to online pornography, about the proliferation of Web sites advocat- Even where filtering companies hired staff ing racial hatred, and about other online ex- to review some Web sites, there were serious pression thought to be offensive or dangerous. problems of subjectivity. The political atti- Congress and the states responded in the late tudes of the filter manufacturers were reflected 1990s with censorship laws, but most of them in their blocking decisions, particularly on were struck down by the courts. Partly as a re- such subjects as homosexuality, human rights, sult, parents, employers, school districts, and and criticism of filtering software. The alterna- other government entities turned to privately tive method, self-rating, did not suffer these manufactured Internet filters. disadvantages, but the great majority of online speakers refused to self-rate their sites. Online In the Communications Decency Act of news organizations, for example, were not 1996, for example, Congress attempted to willing to reduce their content to simplistic block minors from Internet pornography letters or codes through self-rating. by criminalizing virtually all “indecent” or “patently offensive” communications online. Third-party filtering thus became the indus- In response to a 1997 Supreme Court deci- try standard. From early filter companies such sion invalidating the law as a violation of the as SurfWatch and Cyber Patrol, the industry First Amendment, the Clinton Administra- quickly expanded, marketing its products tion began a campaign to encourage Internet to school districts and corporate employ- filtering. ers as well as families. Most of the products contained multiple categories of potentially Early filtering was based on either “self- rating” by online publishers or “third-party Two scholars estimated the size of the World Wide Web in January 2005 at more than 11.5 billion separate index- able pages. A. Gulli & A. Signorini, “The Indexable Web is eno v. ACLU, 521 U.S. 844, 870 (1997), quoting ACLU v. R More Than 11.5 Billion Pages” (May 2005). Source citations Reno, 929 F. Supp. 824, 842 (E.D. Pa. 1996). throughout this report do not include URLs if they can be Id. found in the Bibliography. Brennan Center for Justice offensive or “inappropriate” material. (Some able sites by filtering software to be incom- had more than 50 categories.) Internet service patible with the basic function of libraries, providers such as America Online provided and advocated alternative approaches such as parental control options using the same tech- privacy screens and “acceptable use” policies. nology. Meanwhile, anti-filtering groups such as the Censorware Project and Peacefire began to Some manufacturers marketed products publish reports on the erroneous or question- that were essentially “whitelists” — that is, able blocking of Internet sites by filtering they blocked most of the Internet, leaving just products. a few hundred or thousand pre-selected sites accessible. The more common configuration, In December 2000, President Clinton though, was some form of blacklist, created signed the “Children’s Internet Protection through technology that trolled the Web for Act” (CIPA). CIPA requires all schools and suspect words and phrases. Supplementing the libraries that receive federal financial assis- blacklist might be a mechanism that screened tance for Internet access through the e-rate or Web searches as they happened; then blocked “universal service” program, or through direct those that triggered words or phrases embed- federal funding, to install filters on all com- ded in the company’s software program. puters used by adults as well as minors. The marketing claims of many filtering Technically, CIPA only requires libraries companies were exaggerated, if not flatly false. and schools to have a “technology protec- One company, for example, claimed that its tion measure” that prevents access to “vi- “X-Stop” software identified and blocked only sual depictions” that are “obscene” or “child “illegal” obscenity and child pornography. pornography,” or, for computers accessed by This was literally impossible, since no one minors, depictions that are “obscene,” “child can be sure in advance what a court will rule pornography,” or “harmful to minors.” But “obscene.” The legal definition of obscenity no “technological protection measure” (that is, depends on subjective judgments about “pru- no filter) can make these legal judgments, and rience” and “patent offensiveness” that will be even the narrowest categories offered by filter different for different communities. manufacturers, such as “adult” or “pornog- raphy,” block both text and “visual depic- The “Children’s Internet tions” that almost surely would not be found Protection Act” (CIPA) obscene, child pornography, or “harmful to The late 1990s saw political battles in many minors” by a court of law. communities over computer access in public Public Law 106-554, §1(a)(4), 114 Stat. 2763A-335, amend- libraries. New groups such as Family Friendly ing 20 U.S. Code §6801 (the Elementary & Secondary Edu- Libraries attacked the American Library As- cation Act); 20 U.S. Code §9134(b) (the Museum & Library sociation (ALA) for adhering to a no-censor- Services Act); and 47 U.S. Code §254(h) (the e-rate provision of the Communications Act). ship and no-filtering policy, even for minors. “ Harmful to minors” is a variation on the three-part obscenity The ALA and other champions of intellectual test for adults (see note 4). CIPA defines it as: “any picture, freedom considered the overblocking of valu- image, graphic image file, or other visual depiction that (i) taken as a whole and with respect to minors, appeals to a prurient interest in nudity, sex, or excretion; e Supreme Court defined obscenity for constitutional Th (ii) depicts, describes, or represents, in a patently offensive purposes in Miller v. California, 413 U.S. 15, 24 (1973). The way with respect to what is suitable for minors, an actual or three-part Miller test asks whether the work, taken as a whole, simulated sexual act or sexual contact, actual or simulated lacks “serious literary, artistic, political or scientific value”; normal or perverted sexual acts, or a lewd exhibition of the whether, judged by local community standards, it appeals pri- genitals; and marily to a “prurient” interest; and whether—again judged by (iii) taken as a whole, lacks serious literary, artistic, political, community standards—it describes sexual organs or activities or scientific value as to minors.” in a “patently offensive way.” 47 U.S. Code §254(h)(7)(G). Internet Filters: A Public Policy Report By delegating blocking decisions to pri- A three-judge federal court was convened to vate companies, CIPA thus accomplished far decide the library suit. After extensive fact- broader censorship than could be achieved finding on the operation and performance through a direct government ban. As the of filters, the judges struck down CIPA as evidence in the case that was brought to applied to libraries. They ruled that the law challenge CIPA showed, filters, even when forces librarians to violate their patrons’ First set only to block “adult” or “sexually explicit” Amendment right of access to information content, in fact block tens of thousands of and ideas. nonpornographic sites. The decision included a detailed discus- CIPA does permit library and school sion of how filters operate. Initially, they administrators to disable the required filters trawl the Web in much the same way that “for bona fide research or other lawful pur- search engines do, “harvesting” for possibly poses.” The sections of the law that condition relevant sites by looking for key words and direct federal funding on the installation of phrases. There follows a process of “winnow- filters allow disabling for minors and adults; ing,” which also relies largely on mechanical the section governing the e‑rate program techniques. Large portions of the Web are only permits disabling for adults. never reached by the harvesting and winnow- ing process. CIPA put school and library administra- tors to a difficult choice: forgo federal aid in The court found that most filtering compa- order to preserve full Internet access, or install nies also use some form of human review. But filters in order to keep government grants and because 10,000‑30,000 new Web pages enter e-rate discounts. Not surprisingly, wealthy their “work queues” each day, the companies’ districts were better able to forgo aid than their relatively small staffs (between eight and a lower‑income neighbors. The impact of CIPA few dozen people) can give at most a cursory thus has fallen disproportionately on lower‑in- review to a fraction of the sites that are har- come communities, where many citizens’ only vested, and human error is inevitable. access to the Internet is in public schools and As a result of their keyword-based tech- libraries. CIPA also hurts other demographic nology, the three-judge court found, filters groups that are on the wrong side of the “digi- wrongly block tens of thousands of valuable tal divide” and that depend on libraries for Web pages. Focusing on the three filters used Internet access, including people living in rural most often in libraries — Cyber Patrol, Bess, areas, racial minorities, and the elderly. and SmartFilter — the court gave dozens of In 2001, the ALA, the American Civil examples of overblocking, among them: a Liberties Union, and several state and lo- Knights of Columbus site, misidentified by cal library associations filed suit to challenge Cyber Patrol as “adult/sexually explicit”; a the library provisions of CIPA. No suit was site on fly fishing, misidentified by Bess as brought to challenge the school provisions, “pornography”; a guide to allergies and a site and by 2005, the Department of Education opposing the death penalty, both blocked by estimated that 90% of K-12 schools were Bess as “pornography”; a site for aspiring den- using some sort of filter in accordance with tists, blocked by Cyber Patrol as “adult/sexu- CIPA guidelines. ally explicit”; and a site that sells religious wall hangings, blocked by WebSENSE as “sex.”10 2 0 U.S. Code §6777(c); 20 U.S. Code §9134(f )(3); 47 U.S. Code §254(h)(6)(d). merican Library Association v. United States, 201 F. Supp. 2d A Corey Murray, “Overzealous Filters Hinder Research,” eSchool 401, 431-48 (E.D. Pa. 2002). News Online (Oct. 13, 2005). 10 Id., 431-48. Brennan Center for Justice The judges noted also that filters frequently on the “disabling” provisions of the law as a block all pages on a site, no matter how inno- way for libraries to avoid restricting adults’ cent, based on a “root URL.” The root URLs access to the Internet. Kennedy emphasized for large sites like Yahoo or Geocities contain that if librarians fail to unblock on request, not only educational pages created by non- or adults are otherwise burdened in their profit organizations, but thousands of person- Internet searches, then a lawsuit challenging al Web pages. Likewise, the court found, one CIPA “as applied” to that situation might be item of disapproved content — for example, appropriate.14 a sexuality column on Salon.com — often Three justices—John Paul Stevens, David results in blocking of the entire site.11 Souter, and Ruth Bader Ginsberg—dissented The trial court struck down CIPA’s library from the Supreme Court decision uphold- provisions as applied to both adults and mi- ing CIPA. Their dissents drew attention to nors. It found that there are less burdensome the district court’s detailed description of ways for libraries to address concerns about how filters work, and to the delays and other illegal obscenity on the Internet, and about burdens that make discretionary disabling minors’ access to material that most adults a poor substitute for unfettered Internet ac- consider inappropriate for them — including cess. Souter objected to Rehnquist’s analogy “acceptable use” policies, Internet use logs, between Internet filtering and library book and supervision by library staff.12 selection, arguing that filtering is actually more akin to “buying an encyclopedia and The government appealed the decision then cutting out pages.” Stevens, in a separate of the three-judge court, and in June 2003, dissent, noted that censorship is not necessar- the Supreme Court reversed, upholding the ily constitutional just because it is a condition constitutionality of CIPA. Chief Justice Wil- of government funding—especially when liam Rehnquist’s opinion (for a “plurality” of funded programs are designed to facilitate four of the nine justices) asserted that library free expression, as in universities and libraries, patrons have no right to unfiltered Internet or on the Internet.15 access — that is, filtering is no different, in principle, from librarians’ decisions not to select certain books for library shelves. More- Living with CIPA After the Supreme Court upheld CIPA, pub- over, Rehnquist said, because the government lic libraries confronted a stark choice — forgo is providing financial aid for Internet access, it federal aid, including e-rate discounts, or can limit the scope of the information that is invest resources in a filtering system that, accessed. He added that if erroneous blocking even at its narrowest settings, will censor large of “completely innocuous” sites creates a First quantities of valuable material for reasons Amendment problem, “any such concerns are usually known only to the manufacturer. The dispelled” by CIPA’s provision giving librar- ALA and other groups began developing in- ies the discretion to disable the filter upon formation about different filtering products, request from an adult.13 and suggestions for choosing products and Justices Anthony Kennedy and Stephen settings that block as little of the Internet as Breyer wrote separate opinions concurring in possible, consistent with CIPA. the judgment upholding CIPA. Both relied These materials remind librarians that 11 Id. 14 Id., 2309-12 (concurring opinions of Justices Kennedy and 12 Id., 480-84. Breyer). 13 .S. v. American Library Association, 123 S. Ct. 2297, 2304- U 15 I d., 2317, 2321-22 (dissenting opinions of Justices Stevens 09 (2003) (plurality opinion). and Souter). Internet Filters: A Public Policy Report whatever filter system they choose should in lost e-rate funds, the “community doesn’t allow configuration of the default page to want filtering.”18 educate the user on how the filter works and Likewise, several libraries in New Hampshire how to request disabling. Libraries should decided to forgo federal aid. They were encour- adopt systems that can be easily disabled, in aged by the New Hampshire Library Associa- accordance with the Supreme Court’s state- tion, which posted a statement on its Web ment that CIPA doesn’t violate the First site noting that filters block research on breast Amendment in large part because it autho- cancer, sexually transmitted diseases, “and even rizes librarians to disable filters on the request Super Bowl XXX.”19 of any adult.16 These libraries were the exception, though. In order to avoid commercial products that A preliminary study by the ALA and the maintain secret source codes and blacklists, Center for Democracy and Technology in the Kansas library system developed its own 2004, based on a sample of about 50 librar- filter, KanGuard. Billed as “a library-friendly ies, indicated that a large majority now use alternative,” KanGuard was created by cus- filters, “and most of the filtering is motivated tomizing the open-source filter SquidGuard, by CIPA requirements.” Only 11% of the and aims to block only pornography. But libraries that filter confine their filters to the although KanGuard’s and SquidGuard’s open children’s section. 64% will disable the filter lists may make it easier for administrators to upon request, but fewer than 20% will disable unblock nonpornographic sites that are erro- the filter for minors as well as adults.20 This neously targeted, they cannot avoid the errors picture contrasts sharply with the situation be- of the commercial products, since they rely on fore the Supreme Court’s decision upholding essentially the same technology.17 CIPA, when researchers reported that 73% of libraries overall, and 58% of public libraries, How have libraries responded to CIPA? Ac- did not use filters. 43% of the public libraries cording to reports collected by the ALA, some were receiving e-rate discounts; only 18.9% systems have decided to forgo federal aid or said they would not continue to apply for the e-rate discounts rather than install filters. One e-rate should CIPA be upheld.21 of them, in San Francisco, is subject to a city ordinance that “explicitly bans the filtering of In 2005, the Rhode Island affiliate of the Internet content on adult and teen public ac- American Civil Liberties Union reported cess computers.” A librarian at the San Fran- that before the Supreme Court upheld CIPA, cisco Public Library explained that although fewer than ¼ of the libraries in the state that the ban could cost the library up to $225,000 responded to its survey had installed Inter- net filters, and many had official policies 16 E.g., Lori Bowen Ayre, Filtering and Filter Software (ALA Library Technology Reports, 2004); Open Net Initiative, “Introduction to Internet Filtering” (2004); Derek Hansen, 18 J oseph Anderson, “CIPA and San Francisco: Why We Don’t “CIPA: Which Filtering Software to Use?” (Aug. 31, 2003). Filter,” WebJunction (Aug. 31, 2003). 17 e coordinator of the system says that KanGuard’s lists are Th 19 Associated Press, “Libraries Oppose Internet Filters, Turn compiled “with an open-source ‘robot’ program that scours Down Federal Funds” (June 13, 2004). the Web, searching for language and images that are clearly obscene or harmful to minors.” Walter Minkel, “A Filter 20 Center for Democracy & Technology & ALA, “Children’s That Lets Good Information In,” TechKnowledge (Mar. 1, Internet Protection Act Survey: Executive Summary” (2004) 2004). But no “robot” looking for language or images can (on file at the Free Expression Policy Project). make these legal determinations, and SquidGuard admits 21 Paul Jaeger, John Carlo Bertot, & Charles McClure, “The that its blacklists “are entirely the product of a dumb robot. Effects of the Children’s Internet Protection Act (CIPA) We strongly recommend that you review the lists before in Public Libraries and its Implications for Research: A using them.” “The SquidGuard Blacklist,” www.squidguard. Statistical, Policy, and Legal Analysis,” 55(13) Journal of the org/blacklist (visited 4/3/05). As of 2005, the “porn” section American Society for Information Science and Technology 1131, of SquidGuard had more than 100,000 entries. 1133 (2004). Brennan Center for Justice prohibiting them. By July 1, 2004, how- Public schools also have to deal with the ever—the government’s deadline for imple- complexities and choices occasioned by menting CIPA — all of them were using the CIPA. In 2001, the Consortium for School WebSENSE filter, as recommended by the Networking (CoSN) published a primer, statewide consortium responsible for Internet Safeguarding the Wired Schoolhouse, di- access in libraries.22 rected at policymakers in K-12 schools. The primer seemed to accept filtering as a politi- Although each library system in Rhode cal necessity in many school districts; after Island is allowed to choose its own filter set- Congress passed CIPA, of course, it became tings, the survey showed that most of them a legal necessity as well. CoSN’s later materi- followed the consortium’s recommendations als outline school districts’ options, but note and configured WebSENSE to block the “sex,” that its resources “should not be read as an “adult content,” and “nudity” categories. Four endorsement of [CIPA], of content controls libraries blocked additional categories such as in general, or of a particular technological “gambling,” “games,” “personals and dating,” approach.”25 and “chat.” And even though the Supreme Court conditioned its approval of CIPA on the A further indication of the educational envi- ability of libraries to disable filters on request, ronment came from a reporter who observed: the survey found that many of the state’s Many school technology coordina- library directors were confused about disabling tors argue that the inexact science of their filter and had received no training on Internet filtering and blocking is a how to do so. More than ⅓ of them said they reasonable trade-off for greater peace did not notify patrons that the filters could be of mind. Given the political reality disabled or even that they were in use.23 in many school districts, they say, the The Rhode Island ACLU concluded with choice often comes down to censor- four recommendations on how to minimize ware or no Internet access at all. CIPA’s impact on access to information: He quotes an administrator as saying: “It • Filters should be set at the minimum block- would be politically disastrous for us not to ing level necessary to comply with the law; filter. All the good network infrastructure we’ve installed would come down with the • Libraries should notify patrons that they first instance of an elementary school student have a right to request that the filter be accessing some of the absolutely raunchy sites disabled; out there.”26 • Libraries should train their staff on how to Yet studies indicate that filters in schools disable the filter and on patrons’ right to also frustrate legitimate research and exacer- request disabling; and bate the digital divide.27 The more privileged • All adult patrons should be given the op- portunity to use an unfiltered Internet con- 25 “ School District Options for Providing Access to Appropri- nection.24 ate Internet Content” (power point), www.safewiredschools. org/pubs_and_tools/sws_powerpoint.ppt (visited 2/21/06); 22 Amy Myrick, Reader’s Block: Internet Censorship in Rhode Safeguarding the Wired Schoolhouse (CoSN, June 2001). Island Public Libraries (Rhode Island ACLU, 2005). 26 Lars Kongshem, “Censorware—How Well Does Internet 23 yrick, 16. Moreover, on two occasions when a researcher M Filtering Software Protect Students?” Electronic School Online asked a librarian at the Providence Public library to unblock a (Jan. 1998) (quoting Joe Hill, supervisor at Rockingham wrongly blocked site, the librarian refused and subjected the County, Virginia Public Schools). researcher to judgmental comments and questioning about 27 S ee the reports of the Electronic Frontier Foundation/Online the site’s subject matter. Id., 15. Policy Group and the Kaiser Family Foundation; and the 24 Id., 17. PhD Dissertation of Lynn Sutton, pages 66, 62, 70. Internet Filters: A Public Policy Report students, who have unfiltered Internet access pendix from the 2001 report listing blocked at home, are able to complete their research sites according to subject: artistic and liter- projects. The students from less prosperous ary sites; sexuality education; gay and lesbian homes are further disadvantaged in their edu- information; political topics; and sites relating cational opportunities. to censorship itself. This appendix is available online at www.fepproject.org/policyreports/ Filtering Studies During and appendixa.html. After 2001 Part II describes the tests and studies pub- By 2001, some filter manufacturers said lished during or after 2001. Several of these that they had corrected the problem of are larger and more ambitious than the earlier overblocking, and that instead of keywords, studies, and combine empirical results with they were now using “artificial intelligence.” policy and legal analysis. Our summaries of But no matter how impressive-sounding the these more complex reports are necessarily terminology, the fact remains that all filtering longer than the summaries in the 2001 re- depends on mechanical searches to identify search scan. We have focused on the empirical potentially inappropriate sites. Although some results and sometimes, in the interest of read- of the sillier technologies—such as blocking ability, have rounded off statistics to the near- one word in a sentence and thereby changing est whole number. We urge readers to consult the entire meaning28—are less often seen today, the studies themselves for further detail. studies have continued to document the erro- neous blocking of thousands of valuable Web Some of the reports described in Part II sites, much of it clearly due to mechanical attempt to estimate the overall statistical ac- identification of key words and phrases.29 curacy of different filtering products. Filtering companies sometimes rely on these reports to The first edition of Internet Filters: A Public boast that their percentage of error is relatively Policy Report was intended to advance in- small. But reducing the problems of over- and formed debate on the filtering issue by sum- underblocking to numerical percentages is marizing all of the studies and tests to date, in problematic. one place and in readily accessible form. This second edition brings the report up-to-date, For one thing, percentages and statistics can with summaries of new studies and additional be easily manipulated. Since it is very dif- background on the filtering dilemma. ficult to create a truly random sample of Web sites for testing, the rates of over- and under- Part I is a revision of our 2001 report, and blocking will vary depending on what sites is organized by filtering product. Necessar- are chosen. If, for example, the test sample ily, there is some overlap, since many studies has a large proportion of nonpornographic sampled more than one product. We have up- educational sites on a controversial topic such dated the entries to reflect changes in blocking as birth control, the error rate will likely be categories, or the fact that some of the filters much higher than if the sample has a large mentioned are no longer on the market. In number of sites devoted to children’s toys. the interest of space, we have omitted an ap- Overblocking rates will also vary depending on the denominator of the fraction—that is, 28 e most notorious example was CYBERsitter’s blocking the Th whether the number of wrongly blocked sites word “homosexual” in the phrase: “The Catholic Church opposes homosexual marriage” (see page 22). is compared to the overall total of blocked 29 In addition to the primary research reports described in Part sites or to the overall total of sites tested.30 II, see Commission on Child Online Protection (COPA), Re- port to Congress, (Oct. 20, 2000); National Research Council, 30 S ee the discussion of Resnick et al.’s article on test methodol- Youth, Pornography, and the Internet (2002). ogy, page 45. Brennan Center for Justice Moreover, even when researchers report a Many companies now offer all-purpose relatively low error rate, this does not mean “Web protection” tools that combine censor- that the filter is a good tool for libraries, ship-based filters with other functions such schools, or even homes. With billions of as screening out spam and viruses. Security Internet pages, many changing daily, even a screening tools have become necessary on the 1% error rate can result in millions of wrongly Internet, but they are quite different from blocked sites. filters that block based not on capacity to Finally, there are no consistent or agreed- harm a computer or drown a user’s mailbox upon criteria for measuring over- and un- with spam, but on a particular manufacturer’s derblocking. As we have seen, no filter can concept of offensiveness, appropriateness, or make the legal judgments required by CIPA. child protection. But even if errors are measured based on the content categories created by filter manufac- The Continuing Challenge turers, it is not always easy for researchers to Internet filtering continues to be a major decide whether particular blocked pages fit policy issue, and a challenge for our system within those categories. Percentage summaries of free expression. Some might say that the of correctly or incorrectly blocked sites are debate is over and that despite their many often based on mushy and variable underly- flaws, filters are now a fact of life in American ing judgments about what qualifies as, for homes, schools, offices, and libraries. But example, “alternative lifestyle,” “drug culture,” censorship on such a large scale, controlled by or “intolerance.” private companies that maintain secret black- Because of these difficulties in coming lists and screening technologies, should always up with reliable statistics, and the ease with be a subject of debate and concern. which error rates can be manipulated, we be- lieve that the most useful research on Internet We hope that the revised and updated filters is cumulative and descriptive—that is, Internet Filters will be a useful resource for research that reveals the multitude of sites that policymakers, parents, teachers, librarians, are blocked and the types of information and and all others concerned with the Internet, in- ideas that filters censor from view. tellectual freedom, or the education of youth. Since the first edition of Internet Filters, Internet filtering is popular, despite its unreli- the market for these products has expanded ability, because many parents, political leaders, enormously. In our original research, we and educators feel that the alternative—unfet- found studies that tested one or more of 19 tered Internet access—is even worse. But to different filters. In 2005, we found 133 filter- make these policy choices, it is necessary to ing products. Some of them come in multiple have accurate information about what filters formats for home, school, or business mar- do. Ultimately, as the National Research kets. But there are probably fewer than 133 Council observed in a 2002 report, less censo- separate products, because the basic software rial approaches such as media literacy and for popular filters like Bess and Cyber Patrol sexuality education are the only effective ways is licensed to Internet service providers and to address concerns about young people’s ac- other companies that want to offer filtering cess to controversial or disturbing ideas.31 under their own brand name. 31 National Research Council, Youth, Pornography, and the Internet, Exec. Summary; ch. 10. Internet Filters: A Public Policy Report I.The 2001 Research Scan Updated: Over- and Underblocking by Internet Filters This section is organized by filtering product, the defects of various filtering products with- and describes each test or study that we found out identifying particular blocked sites. Access up through the fall of 2001. Denied, Version 2.0 addressed AOL Parental Controls only in its introduction, where it America Online Parental Controls reported that the “Kids Only” setting blocked AOL offers three levels of Parental Controls: the Web site of Children of Lesbians and Gays “Kids Only,” for children 12 and under; Everywhere (COLAGE), as well as a number “Young Teen,” for ages 13-15; and “Mature of “family, youth and national organization Teen,” for ages 16-17, which allows access Web sites with lesbian and gay content,” none to “all content on AOL and the Internet, of which were named in the report. except certain sites deemed for an adult (18+) Brian Livingston, “AOL’s ‘Youth Filters’ audience.”32 AOL encourages parents to cre- Protect Kids From Democrats,” CNet News ate unique screen names for their children (Apr. 24, 2000) and to assign each name to one of the four age categories. At one time, AOL employed Livingston investigated AOL’s filtering for Cyber Patrol’s block list; at another point it signs of political bias. He found that the “Kids stated it was using SurfWatch. In May 2001, Only” setting blocked the Web sites of the AOL announced that Parental Controls had Democratic National Committee, the Green integrated the RuleSpace Company’s “Contex- Party, and Ross Perot’s Reform Party, but not ion Services,” which identifies “objectionable” those of the Republican National Committee sites “by analyzing both the words on a page and the conservative Constitution and Lib- and the context in which they are used.”33 ertarian parties. AOL’s “Young Teen” setting blocked the home pages of the Coalition to Gay and Lesbian Alliance Against Defama- Stop Gun Violence, Safer Guns Now, and the tion (GLAAD), Access Denied, Version 2.0: Million Mom March, but neither the Nation- The Continuing Threat Against Internet Access al Rifle Association site nor the commercial and Privacy and Its Impact on the Lesbian, Gay, sites for Colt & Browning firearms. Bisexual and Transgender Community (1999) Bennett Haselton, “AOL Parental Controls This 1999 report was a follow-up to Error Rate for the First 1,000 .com Domains” GLAAD’s 1997 publication, Access Denied: (Peacefire, Oct. 23, 2000) The Impact of Internet Filtering Software on the Lesbian and Gay Community, which described Peacefire Webmaster Bennett Haselton tested AOL Parental Controls on 1,000 dot- 32 AOL, “Parental Controls,” site.aol.com/info/parentcontrol. com domains he had compiled for a similar html (visited 3/6/06). 33 AOL press release, “AOL Deploys RuleSpace Technology test of SurfWatch two months earlier (see page Within Parental Controls” (May 2, 2001), www.rulespace. 36). He attempted to access each site on AOL com/news/pr107.php (visited 2/23/06). Brennan Center for Justice 5.0 adjusted to its “Mature Teen” setting. Five Bess of the 1,000 working domains were blocked, Bess, originally manufactured by N2H2, including a-aji.com, a site that sold vinegar was acquired by Secure Computing in and seasonings. Haselton decided the four October 2003. By late 2005, Bess had been others were pornographic and thus accurately merged into SmartFilter, another Secure blocked. This produced an “error rate” of Computing product, and was being marketed 20%, the lowest, by Peacefire’s calculation, of to schools under the name SmartFilter, Bess the five filters tested. AOL also “blocked far Edition.34 fewer pornographic sites than any of the other programs,” however. Haselton stated that five Bess combines technology with some hu- blocked domains was an insufficient sample to man review. Although N2H2 initially claimed gauge the efficacy of AOL Parental Controls that all sites were reviewed by its employees accurately, and that the true error rate could before being added to the block list, the cur- fall anywhere between 5-75%. rent promotional literature simply states that the filter’s “unique combination of technology “Digital Chaperones for Kids,” Consumer and human review … reduces frustrations as- Reports (Mar. 2001) sociated with ‘keyword blocking’ methods, in- Consumer Reports assessed AOL’s “Young cluding denied access to sites regarding breast Teen” and “Mature Teen” settings along with cancer, sex education, religion, and health.”35 various other filtering technologies. For each In 2001, Bess had 29 blocking categories; filter, the researchers attempted to access by 2006, the number was 38, ranging from 86 Web sites that they deemed objection- “adults only” and “alcohol” to “gambling,” able because of “sexually explicit content or “jokes,” “lingerie,” and “tasteless/gross.” Its violently graphic images,” or promotion of four “exception” categories in 2001 were ex- “drugs, tobacco, crime, or bigotry.” They also panded to six: “history,” “medical,” “moderat- tested the filters against 53 sites they deemed ed,” “text/spoken only,” “education,” and “for legitimate because they “featured serious con- kids.” Each exception category allows access tent on controversial subjects.” The “Mature to sites that have educational value but might Teen” setting left 30% of the “objectionable” otherwise be filtered – for example, children’s sites unblocked; the “Young Teen” filter failed games that would be blocked under “games” to block 14% – the lowest underblocking rate or “jokes”; classic literature, history, art, or sex of all products reviewed. But “Young Teen” education that would be blocked under “sex,” also blocked 63% of the “legitimate” sites, “nudity,” or “violence.” including Peacefire.org; Lesbian.org, an online guide to lesbian politics, history, arts, and Karen Schneider, A Practical Guide to Internet culture; the Citizens’ Committee for the Right Filters (1997) to Keep and Bear Arms; the Southern Poverty From April to September 1997, Karen Law Center; and SEX, Etc., a sex education Schneider supervised a nationwide team of site hosted by Rutgers University. librarians in testing 13 filters, including Bess. Miscellaneous Reports 34 “ Secure Computing Acquires N2H2,” www.securecomput- ing.com/index.cfm?skey=1453 (visited 3/3/06). Secure • In “BabelFish Blocked by Censorware” Computing also embeds filtering for “inappropriate” content (Feb. 27, 2001), Peacefire reported that in other pr.oducts such as CyberGuard and Webwasher. “Secure Computing Products at a Glance,” www.bess. AOL’s “Mature Teen” setting barred access com/index.cfm?skey=496; www.securecomputing.com/index. to BabelFish, AltaVista’s foreign-language cfm?skey=496 (visited 3/3/06). translation service. 35 “ SmartFilter, Bess Edition, Filtering Categories,” www.bess. com/index.cfm?skey=1379 (visited 3/3/06). 10 Internet Filters: A Public Policy Report The results of this Internet Filter Assessment neither the names nor the Web addresses of Project, or TIFAP, were published later that the blocked sites. year in Schneider’s Practical Guide to Internet Censorware Project, Passing Porn, Banning the Filters. Bible: N2H2’s Bess in Public Schools (2000) The researchers began by seeking answers From July 23-26, 2000, the Censorware to 100 common research queries, on both Project tested “thousands” of URLs against unfiltered computers and ones equipped with 10 Bess proxy servers, seven of which were in Bess (and the various other filters) configured use in public schools across the United States. for maximum blocking, including keyword Among the blocked sites were a page from blocking. Each query fell into one of 11 Mother Jones magazine; the Institute of Aus- categories: “sex and pornography,’” “anatomy,” tralasian Psychiatry; the nonprofit effort Stop “drugs, alcohol, and tobacco,” “gay issues,” Prisoner Rape; and a portion of the Columbia “crimes (including pedophilia and child por- University Health Education Program site, on nography),” “obscene or ‘racy’ language,” “cul- which users are invited to submit “questions ture and religion,” “women’s issues,” “gam- about relationships; sexuality: sexual health; bling,” “hate groups and intolerance,” and emotional health; fitness; nutrition; alcohol, “politics.” The queries were devised to gauge nicotine, and other drugs; and general health.” filters’ handling of controversial issues – for Bess also blocked the United Kingdom-based instance, “I’d like some information on safe Feminists Against Censorship, the personal sex”; “I want information on the legalization site of a librarian opposing Internet filter of marijuana”; “Is the Aryan Nation the same use in libraries, and Time magazine’s “Netly thing as Nazis?” and “Who are the founders News,” which had reported, positively and of the Electronic Frontier Foundation and negatively, on filtering software. what does it stand for?” In some cases, the queries contained potentially provocative The report noted that, contrary to the im- terms “intended to trip up keyword-blocking plication in Bess’s published filtering criteria, mechanisms,” such as “How do beavers make Bess does not review home pages hosted by their dams?”; “Can you find me some pictures such free site providers as Angelfire, Geocities, from Babes in Toyland?”; and “I’m trying to and Tripod (owing, it seems, to their sheer find out about the Paul Newman movie The number). Instead, users must configure the Hustler.” software to block none or all of these sites; some schools opt for the latter, thus prohibit- Schneider used Web sites, blocked and ing access to such sites as The Jefferson Bible, a unblocked, that arose from these searches to compendium of Biblical passages selected by construct a test sample of 240 URLs. Her Thomas Jefferson, and the Eustis Panthers, a researchers tested these URLs against a version high school baseball team. Though each proxy of Bess configured for “maximum filtering,” was configured to filter out pornography to but with keyword filtering disabled. TIFAP the highest degree, Censorware reported that found that “several” (Schneider did not say it was able to access hundreds of pornographic how many) nonpornographic sites were Web sites, of which 46 are listed in Passing blocked, including a page discussing X-rated Porn. videos but not containing any pornographic imagery, and an informational page on tri- Peacefire, “‘BESS, the Internet Retriever’ chomaniasis, a vaginal disease. Upon notifi- Examined” (2000; since updated) cation and review, Bess later unblocked the This report lists 15 sites that Peacefire trichomaniasis site. A Practical Guide included deemed inappropriately blocked by Bess Brennan Center for Justice 11 during the first half of 2000. They included weekend; the home page of “American Gov- Peacefire.org itself, which was blocked for ernment and Politics,” a course at St. John’s “Profanity” when the word “piss” appeared on University; and the Circumcision Information the site (in a quotation from a letter written and Research Pages, a site that contained no by Brian Milburn, president of CYBERsitter’s nudity and was designated a “Select Parenting manufacturer, Solid Oak Software, to jour- Site” by ParenthoodWeb.com. nalist Brock Meeks). Also blocked were: two Bennett Haselton, “BESS Error Rate for 1,000 portions of the Web site of Princeton Univer- .com Domains” (Peacefire, Oct. 23, 2000) sity’s Office of Population Research; the Safer Sex page; five gay-interest sites, including the Bennett Haselton performed the same test home page of the Illinois Federation for Hu- of 1,000 active dot-com domains for Bess as man Rights; two online magazines devoted to he did for AOL (see page 9). N2H2 officials gay topics; two Web sites providing resources had evidently reviewed his earlier report on on eating disorders; and three sites discussing SurfWatch, and prepared for a similar test by breast cancer.36 unblocking any of the 1,000 sites inappropri- ately filtered by Bess,37 so Peacefire selected Jamie McCarthy, “Mandated Mediocrity: a second 1,000 dot-com domains for testing Blocking Software Gets a Failing Grade” against a Bess proxy server in use at a school (Peacefire/ Electronic Privacy Information where a student had offered to help measure Center, Oct. 2000) Bess’s performance. “Mandated Mediocrity” describes another The filter was configured to block sites 23 Web sites inappropriately blocked by Bess. in the categories of “adults only,” “alcohol,” The URLs were tested against an N2H2 proxy “chat,” “drugs,” “free pages,” “gambling,” as well as a trial copy of the N2H2 Inter- “hate/discrimination,” “illegal,” “lingerie,” net Filtering Manager set to “typical school “nudity,” “personals,” “personal information,” filtering.” Among the blocked sites were the “porn site,” “profanity,” “school cheating Traditional Values Coalition; “Hillary for info,” “sex,” “suicide/murder,” “tasteless/gross,” President”; “The Smoking Gun,” an online “tobacco,” “violence,” and “weapons.” The Bess blocked the Traditional keyword-blocking features were also enabled. The BESS proxy blocked 176 of the 1,000 Values Coalition and “Hillary domains; among these, 150 were “under for President.” construction.” Of the remaining 26 sites, Peacefire deemed seven wrongly blocked: selection of primary documents relating to a-celebrity.com, a-csecurite.com, a-desk.com, current events; a selection of photographs of a-eda.com, a-gordon.com, a-h-e.com, and Utah’s national parks; “What Is Memorial a-intec.com. Day?”, an essay lamenting the “capitalistic American” conception of the holiday as noth- The report said the resulting “error rate” of ing more than an occasion for a three-day 27% was unreliable given how small a sample was examined; the true error rate “could be 36 ese last three pages were not filtered because of an auto- Th as low as 15%.” Haselton also noted that the matic ban on the keyword “breast,” but either were reviewed dot-com domains tested here were “more and deemed unacceptable by a Bess employee, or had other words or phrases that triggered the filter. The report noted: likely to contain commercial pornography “In our tests, we created empty pages that contained the than, say, .org domains. ... We should expect words breast and breast cancer in the titles, to test whether Bess was using a word filter. The pages we created were ac- cessible, but the previous three sites about breast cancer were 37 ennett Haselton, “Study of Average Error Rates for Censor- B still blocked.” ware Programs” (Peacefire, Oct. 23, 2000). 12 Internet Filters: A Public Policy Report the error rate to be even higher for .org sites.” pornography, and/or violence.” “While block- He added that the results called into question ing software companies often justify their N2H2 CEO Peter Nickerson’s claim, in 1998 errors by pointing out that they are quickly testimony before a congressional committee, corrected,” the report concluded, “this does that “all sites that are blocked are reviewed by not help any of the candidates listed above. ... N2H2 staff before being added to the block corrections made after Election Day do not lists.”38 help them at all.” Bennett Haselton & Jamie McCarthy, “Blind Bennett Haselton, “Amnesty Intercepted: Ballots: Web Sites of U.S. Political Candidates Global Human Rights Groups Blocked by Censored by Censorware” (Peacefire, Nov. 7, Web Censoring Software” (Peacefire, Dec. 12, 2000) 2000) “Blind Ballots” was published on Election In response to complaints from students Day, 2000. The authors obtained a random barred from the Amnesty International Web sample of political candidates’ Web sites from page, among others, at their school computer NetElection.org, and set out to see which stations, Peacefire examined various filters’ sites Bess’s (and Cyber Patrol’s) “typical school treatment of human rights sites. It found that filtering” would allow users to access. (Around Bess’s “typical school filtering” blocked the the start of the 2000 school year, Bess and home pages of the International Coptic Con- Cyber Patrol asserted that together they were gress, which tracked human rights violations providing filtered Internet access to more than against Coptic Christians living in Egypt; 30,000 schools nationwide.39) and Friends of Sean Sellers, which contained links to the works of the Multiple Personality Bess’s wholesale blocking of free Web host- Disorder-afflicted writer who was executed ing services caused the sites of one Democrat- in 1999 for murders he had committed as a ic candidate, five Republicans, six Libertarians 16-year-old. (The site opposed capital punish- (as well as the entire Missouri Libertarian ment.) Party site), and 13 other third-party candi- dates to be blocked. The authors commented “Typical school filtering” also denied access that, as “many of our political candidates run to the official sites of recording artists Suzanne their campaigns on a shoestring, and use free- Vega and the Art Dogs; both contained state- hosting services to save money,” Bess’s barring ments that portions of their proceeds would of such hosts leads it to an inadvertent bias be donated to Amnesty International. Bess’s toward wealthy or established politicians’ sites. “minimal filtering” configuration blocked the Congressman Edward Markey (a Democrat Web sites of Human Rights & Tamil People, from Massachusetts), also had his site blocked which tracks government and police violence – unlike the others, it was not hosted by against Hindu Tamils in Sri Lanka; and Casa Geocities or Tripod, but was blocked because Alianza, which documents the condition of Bess categorized its content as “hate, illegal, homeless children in the cities of Central America. 38 Peter Nickerson Testimony, House Subcom. on Telecom- munications, Trade, and Consumer Protection (Sept. 11, Miscellaneous Reports 1998), www.peacefire.org/censorware/BESS/peter-nickerson. filtering-bill-testimony.9-11-1998.txt (visited 3/6/06). • In its “Winners of the Foil the Filter Con- 39 N2H2 press release, “N2H2 Launches Online Curriculum test” (Sept. 28, 2000), the Digital Freedom Partners Program, Offers Leading Education Publishers Access Network reported that Bess blocked House to Massive User Base” (Sept. 6, 2000); Surf Control press release, “Cyber Patrol Tells COPA Commission that Market Majority Leader Richard “Dick” Armey’s for Internet Filtering Software to Protect Kids is Booming” official Web site upon detecting the word (July 20, 2000). Brennan Center for Justice 13 “dick.” Armey, himself a filtering advocate, home page of cyberlaw scholar Lawrence won “The Poetic Justice Award – for those Lessig, who was to testify before the COPA bitten by their own snake.” Commission, Peacefire attempted to access various pages on the COPA Commission site, • Peacefire reported, in “BabelFish Blocked as well as the Web sites of organizations and by Censorware” (Feb. 27, 2001), that Bess companies with which the commissioners blocked the URL-translation site BabelFish. were affiliated, through a computer equipped • In “Teen Health Sites Praised in Article, with ClickSafe. On the Commission’s site, Blocked by Censorware” (Mar. 23, 2001), ClickSafe blocked the Frequently Asked Peacefire’s Bennett Haselton noted that Questions page; the biographies of Commis- Bess blocked portions of TeenGrowth, a sion members Stephen Balkam, Donna Rice teen-oriented health education site that Hughes, and John Bastian; a list of “tech- was recognized by the New York Times in nologies and methods” within the scope of the recent article, “Teenagers Find Health the Commission’s inquiry; the Commission’s Answers with a Click.”40 Scope and Timeline Proposal; and two ver- sions of the COPA law. ClickSafe As for groups with representatives on the Rather than relying on a list of objection- Commission, ClickSafe blocked several orga- able URLs, ClickSafe software reviewed each nizations’ and companies’ sites, at least partial- requested page in real time. In an outline ly: Network Solutions; the Internet Content for testimony submitted to the commission Rating Association; Security Software’s infor- created by the 1998 Child Online Protection mation page on its signature filtering product, Act (the “COPA Commission”), company Cyber Sentinel; FamilyConnect, a brand of co-founder Richard Schwartz claimed that blocking software; the National Law Center ClickSafe “uses state-of-the-art, content-based for Children and Families; the Christian site filtering software that combines cutting edge Crosswalk.com; and the Center for Democ- graphic, word and phrase-recognition technol- racy and Technology (CDT). In addition to ogy to achieve extraordinarily high rates of ac- the CDT, ClickSafe blocked the home pages curacy in filtering pornographic content,” and of the ACLU, the Electronic Frontier Founda- “can precisely distinguish between appropriate tion, and the American Family Association, as and inappropriate sites.”41 This was vigorously well as part of the official site for Donna Rice disputed by Peacefire (see below). By 2005, Hughes’s book, Kids Online: Protecting Your a Web site for the ClickSafe filter could no Children in Cyberspace. longer be found, although a European com- pany using the same name had launched a site focused on Internet safety for minors.42 Cyber Patrol In 2001, Cyber Patrol operated with 12 Peacefire, “Sites Blocked by ClickSafe” (July default blocking categories, including “partial 2000) nudity,” “intolerance,” “drugs/drug culture,” and “sex education.”43 The manufacturer’s Upon learning that ClickSafe blocked the Web site in 2001 implied that “a team of 40 Bonnie Rothman Morris, “Teenagers Find Health Answers with a professional researchers” reviewed all sites to Click,” New York Times (Mar. 20, 2001), F8. decide whether they should be blocked; by 41 Outline for Testimony Presented by Richard Schwartz, Co- Founder, ClickSafe.com, www.copacommission.org/meetings/ 2006, the company described its filter as a mix hearing2/schwartz.test.pdf (visited 3/13/05). 42 “New Clicksafe” (site in Dutch and French), www.clicksafe.be/taalkeuze. 43 By 2005, Cyber Patrol had 13 categories, several of them different html (visited 3/13/05); “Background Clicksafe,” www.saferinternet. from the original 12. CyberList, www.cyberpatrol.com/Default. org/ww/en/pub/insafe/focus/belgium/be_node.htm (visited 3/13/05). aspx?id=123&mnuid=2.5 (visited 3/14/06). 14 Internet Filters: A Public Policy Report of human researchers and automated tools.44 Centers for Disease Control and Prevention, Like most filter manufacturers, Cyber Patrol the AIDS Book Review Journal, and AIDS does not make its list of prohibited sites pub- Treatment News. Cyber Patrol also blocked a lic, but its “test-a-site” search engine (formerly number of newsgroups dealing with homosex- called “CyberNOT”) allows users to enter uality or gender issues, such as alt.journalism. URLs and learn immediately whether those gay-press; soc.support.youth.gay-lesbian-bi; pages are on the list. In 2001, the company alt.feminism; and soc.support.fat-acceptance. stated that it blocked all Internet sites “that Karen Schneider, A Practical Guide to Internet contain information or software programs Filters (1997) designed to hack into filtering software” in all of its blocking categories; this statement is no The Internet Filter Assessment Project tested longer on the Cyber Patrol site. Cyber Patrol configured to block only “full nudity” and “sexual acts.” Schneider reported Brock Meeks & Declan McCullagh, “Jack- that the software “blocked “good sites” 5-10% ing in From the ‘Keys to the Kingdom’ Port,” of the time, and pornographic sites slipped CyberWire Dispatch (July 3, 1996) through about 10% of the time. One of the The first evaluation of Cyber Patrol ap- “good sites” was www.disinfo.com, described peared in this early report on the problems by Schneider as a site “devoted to debunking of Internet filtering by journalists Brock propaganda.” Meeks and Declan McCullagh. They viewed a Jonathan Wallace, “Cyber Patrol: The Friendly decrypted version of Cyber Patrol’s block list Censor” (Censorware Project, Nov. 22, 1997) (along with those of CYBERsitter and Net Nanny), and noticed that Cyber Patrol stored Jonathan Wallace tested approximately 270 the Web addresses it blocked only partially, sites on ethics, politics, and law – all “con- cutting off all but the first three characters at taining controversial speech but no obscenity the end of a URL. For instance, the software or illegal material” – against the CyberNOT was meant to block loiosh.andrew.cmu. search engine after learning that the Web edu/~shawn, a Carnegie Mellon student home pages of Sex, Laws, and Cyberspace, the 1996 page containing information on the occult; book he co-authored with Mark Mangan, yet on its block list Cyber Patrol recorded were blocked by Cyber Patrol. Wallace found only loiosh.andrew.cmu.edu/ ~sha, thereby 12 of his chosen sites were barred, including blocking every site beginning with that Deja News, a searchable archive of Usenet ma- URL segment and leaving, at the time of the terials, and the Society for the Promotion of report’s publication, 23 unrelated sites on the Unconditional Relationships, an organization university’s server blocked. advocating monogamy. He could not find out which of Cyber Patrol’s categories these sites The authors also found that with all de- fit into. When asked, a Cyber Patrol represen- fault categories enabled, Cyber Patrol barred tative simply said that the company consid- multiple sites concerning cyberliberties – the ered the sites “inappropriate for children.” Electronic Frontier Foundation’s censorship archive, for example, and MIT’s League for Wallace reported that Cyber Patrol also Programming Freedom. Also blocked were blocked sites featuring politically loaded the Queer Resources Directory, which counts content, such as the Flag Burning Page, which among its resources information from the examines the issue of flag burning from a con- stitutional perspective; Interactivism, which 44 Cyber Patrol, “Accurate, Current & Relevant Filtering,” www. invites users to engage in political activism cyberpatrol.com/Default.aspx?id=129&mnuid=2.5 (visited 2/26/06). by corresponding with politicians on issues Brennan Center for Justice 15 such as campaign finance reform and Tibetan ered wrongly blocked in the “full nudity” independence; Newtwatch, a Democratic and “sexual acts” categories, among them Party-funded page that consisted of reports Creature’s Comfort Pet Service; Air Penny (a and satires on the former Speaker of the Nike site devoted to basketball player Penny House; Dr. Bonzo, which featured “satirical Hardaway); the MIT Project on Mathematics essays on religious matters”45; and the Second and Computation; AAA Wholesale Nutrition; Amendment Foundation – though, as Wal- the National Academy of Clinical Biochemis- lace noted, Cyber Patrol did not block other try; the online edition of Explore Underwater gun-related sites, such as the National Rifle magazine; the computer science department Association’s. of England’s Queen Mary and Westfield Col- lege; and the United States Army Corps of Gay and Lesbian Alliance Against Defamation Engineers Construction Engineering Research (GLAAD) press release, “Gay Sites Netted in Laboratories. The report took its title from Cyber Patrol Sting” (Dec. 19, 1997) two additional sites blocked for “full nudity” GLAAD reported that Cyber Patrol was and “sexual acts”: “We, the People of Ada,” an blocking the entire “WestHollywood” subdi- Ada, Michigan, committee devoted to “bring- rectory of Geocities. WestHollywood, at that ing about a change for a more honest, fiscally time, was home to more than 20,000 gay- and responsible and knowledgeable township gov- lesbian-interest sites, including the National ernment,” and Yoyo, a server of Melbourne, Black Lesbian and Gay Leadership Forum’s Australia’s Monash University. Young Adult Program. When contacted, Blacklisted also reported that every site Cyber Patrol’s then-manufacturer Microsys- hosted by the free Web page provider Tri- tems Software cited, by way of explanation, pod was barred, not only for nudity or the high potential for WestHollywood sites sexually explicit content, but also for “vio- to contain nudity or pornographic imagery. lence/profanity,” “gross depictions,” “intoler- GLAAD’s press release pointed out, however, ance,” “satanic/cult,” “drugs/drug culture,” that Geocities expressly prohibited “nudity “militant/extreme,” “questionable/illegal & and pornographic material of any kind” on its gambling,” and “alcohol & tobacco.” Tripod server. was home, at the time of the report, to 1.4 Microsystems CEO Dick Gorgens respond- million distinct pages, but smaller servers and ed to further inquiry with the admission that service providers were also blocked in their GLAAD was “absolutely correct in [its] assess- entirety—Blacklisted lists 40 of them. Another ment that the subdirectory block on WestHol- section of the report lists hundreds of blocked lywood is prejudicial to the Gay and Lesbian newsgroups, including alt.atheism; alt.adop- Geocities community. ... Over the next week tion; alt.censorship; alt.journalism; rec.games. the problem will be corrected.” Yet according bridge (for bridge enthusiasts); and support. to the press release, after a week had passed, soc.depression.misc (on depression and mood the block on WestHollywood remained. disorders). Censorware Project, Blacklisted by Cyber Pa- The day after Blacklisted was published, trol: From Ada to Yoyo (Dec. 22, 1997) Microsystems Software unblocked 55 of the 67 URLs and domains the report had cited. This report documented a number of Eight of the remaining 12, according to sites that the Censorware Project consid- the Censorware Project, were still wrongly 45 Wallace added that the blocking of this site, “long removed blocked: Nike’s Penny Hardaway site; the from the Web, raises questions about the frequency with National Academy of Biochemistry sites; which the Cyber Patrol database is updated.” 16 Internet Filters: A Public Policy Report four Internet service providers; Tripod; and a tions in ACLU v. Reno and ACLU v. Reno II, site-in-progress for a software company. This the American Civil Liberties Union’s chal- last site, at the time of Censorware’s Decem- lenges to the 1997 Communications Decency ber 25, 1997 update to Blacklisted, contained Act and 1998 Child Online Protection Act. very little content, but did contain the words Hunter then conducted Yahoo searches for “HOT WEB LINKS,” which was “apparently sites pertaining to Internet portals, political enough for Cyber Patrol to continue to block issues, feminism, hate speech, gambling, re- it as pornography through a second review.” ligion, gay pride and homosexuality, alcohol, Of the four other sites left blocked, two, tobacco, and drugs, pornography, news, vio- Censorware acknowledged, fell within Cyber lent computer games, safe sex, and abortion. Patrol’s blocking criteria and “shouldn’t have From each of the first 12 of these 13 searches, been listed as wrongful blocks originally.”46 Hunter chose five of the resulting matches for his sample, and then selected four abortion- Christopher Hunter, Filtering the Future?: related sites (two pro- and two anti-) in order Software Filters, Porn, PICS, and the Internet to arrive at a total of 100 URLs. Content Conundrum (Master’s thesis, Annen- berg School for Communication, University Hunter evaluated the first page of each site of Pennsylvania, July 1999) using the rating system devised by an indus- try group called the Recreational Software In June 1999, Christopher Hunter tested Advisory Council (RSAC). Under the RSAC’s 200 URLs against Cyber Patrol and three four categories (violence, nudity, sex, and lan- other filters. Contending that existing reports guage) and five grades within each category, a on blocked sites applied “largely unscientific site with a rating of zero in the “sex” category, methods” (that is, they did not attempt to for example, would contain no sexual content assess overall percentages of wrongly blocked or else only “innocent kissing; romance,” sites), Hunter tested Cyber Patrol, CYBERsit- while a site with a “sex” rating of 4 might ter, Net Nanny, and SurfWatch by “social sci- contain “explicit sexual acts or sex crimes.” ence methods of randomization and content Using these categories, Hunter made his own analysis.” judgments as to whether a filtering product Hunter intended half of his testing sample erroneously blocked or failed to block a site, to approximate an average Internet user’s surf- characterizing a site whose highest RSAC ing habits. Thus, the first 100 sites consisted rating he thought would be zero or one as of 50 that were “randomly generated” by nonobjectionable, while determining that any Webcrawler’s random links feature and 50 site with a rating of 2, 3, or 4 in at least one others that Hunter compiled through Alta- RSAC category should have been blocked.47 Vista searches for the five most frequently After testing each filter at its “default” set- requested search terms as of April 1999: ting, Hunter concluded that Cyber Patrol “yahoo,” “warez” (commercial software ob- blocked 20 sites, or 55.6%, of the material tainable for download), “hotmail,” “sex,” and he deemed objectionable according to RSAC “MP3.” Hunter gathered the first 10 matches from each of these five searches. 47 ecause the RSAC’s system depended on self-rating, it B never gained much traction in the U.S., where third-party For the other 100 sites, Hunter focused on filtering products soon dominated the market. In 1999, the RSAC merged with the Internet Content Rating Association material often identified as controversial, such (ICRA), a British-based industry group. See www.rsac.org; as the Web sites of the 36 plaintiff organiza- www.icra.org/about (both visited 3/14/05). For background on RSAC, ICRA, and their difficulty achieving wide ac- ceptance, see Marjorie Heins, Not in Front of the Children: 46 Censorware Project, Blacklisted By Cyber Patrol: From Ada to “Indecency,” Censorship, and the Innocence of Youth (2001), Yoyo – The Aftermath (Dec. 25, 1997). 224, 261, 351-52. Brennan Center for Justice 17 standards, and 15 sites, or 9.1%, of the mate- percentages of wrongly blocked sites, Hunter rial he deemed innocuous. Among the 15 in- relied on his own subjective judgments on how nocuous sites were the feminist literary group the different Web sites fit into the RSAC’s 20 RiotGrrl; Stop Prisoner Rape; the Qworld separate rating categories.49 contents page, a collection of links to online Center for Media Education (CME), Youth gay-interest resources; an article on “Promot- Access to Alcohol and Tobacco Web Marketing: ing with Pride” on the Queer Living page; the The Filtering and Rating Debate (Oct. 1999) Coalition for Positive Sexuality, which pro- motes “complete and honest sex education”; The CME tested Cyber Patrol and five SIECUS, the Sexuality Information and Edu- other filters for underinclusive blocking of cation Council of the United States; and Gay alcohol and tobacco marketing materials. Wired Presents Wildcat Press, a page devoted They first selected the official sites of 10 beer to an award-winning independent press. manufacturers and 10 liquor companies that are popular and “[have] elements that ap- Although Hunter may well have been right peal to youth.” They added 10 sites pertain- that many of the blocked sites were relatively ing to alcohol – discussing drinking games unobjectionable according to the RSAC rat- or containing cocktail-making instructions, ings, Cyber Patrol’s default settings for these for example – and 14 sites promoting smok- filters (for example, “sex education”) were ing. (As major U.S. cigarette brands are not specifically designed to sweep broadly across advertised online, CME chose the home pages many useful sites. It’s not entirely accurate, of such magazines as Cigar Aficionado and therefore, to conclude that the blocking of all Smoke.) Cyber Patrol blocked only 43% of the these sites would be erroneous; rather, it would promotional sites. be the result of restrictive default settings and user failure to disable the pre-set categories. The CME also conducted Web searches on Five of the sites Hunter deemed overblocked three popular search engines – Yahoo, Go/ by Cyber Patrol, for example, were alcohol- InfoSeek, and Excite – for the alcohol- and and tobacco-related, and thus fell squarely tobacco-related terms “beer,” “Budweiser liz- within the company’s default filtering criteria. ards,” “cigarettes,” “cigars,” “drinking games,” “home brewing,” “Joe Camel,” “liquor,” and In February 2000, filtering advocate David “mixed drinks.” It then attempted to access Burt (later to become an employee of N2H2) the first five sites returned in each search. responded to Hunter’s study with a press Cyber Patrol blocked 30% of the result pages, release citing potential sources of error.48 allowing, for example, cigarettes4u.com, Burt argued that “200 sites is far too small to tobaccotraders.com, and homebrewshop.com, adequately represent the breadth of the entire which, according to the report, “not only World Wide Web” and charged that all but the promoted the use of alcohol and tobacco, but 50 randomly generated URLs constituted a also sold products and accessories related to skewed sample, containing content “instantly their consumption.” recognizable as likely to trigger filters” and “not represented in the sample proportion- To test blocking of educational and public ately to the entire Internet,” thus giving rise health information on alcohol and tobacco, to “much higher-than-normal error rates.” 49 Hunter later testified as an expert witness for the plaintiffs in A more serious problem, however, is that in the lawsuit challenging CIPA. The district court noted that attempting to arrive at “scientific” estimates of his attempt to calculate over- and underblocking rates scien- tifically, like a similar attempt by experts for the government, was flawed because neither began with a truly random sample 48 Filtering Facts press release, “ALA Touts Filter Study Whose of Web sites for testing. American Library Ass’n v. U.S., 201 F. Own Author Calls Flawed” (Feb. 18, 2000). Supp. 2d at 437-38. 18 Internet Filters: A Public Policy Report the CME added to its sample 10 sites relating ington State’s Tri-City Herald; part of the to alcohol consumption – for instance, www. City of Hiroshima site; the former Web site alcoholismhelp.com, Mothers Against Drunk of the American Airpower Heritage Museum Driving and the National Organization in Midland, Texas; an Illinois Mathematics on Fetal Alcohol Syndrome, along with 10 and Science Academy student’s personal home anti-smoking sites, and the American Cancer page, which at the time of Jansson and Skala’s Society. Cyber Patrol did not block any of the report consisted only of the student’s résumé; sites in this group. Nor did it block most sites and the Web site of a sheet-music publisher. returned by the three search engines when The “Marston Family Home Page,” a per- terms like “alcohol,” “alcoholism,” “fetal alco- sonal site, was also blocked under the “mili- hol syndrome,” “lung cancer,” or “substance tant/extremist” and “questionable/illegal & abuse” were entered. Cyber Patrol allowed ac- gambling” categories – presumably, according cess to an average of 4.8 of the top five search to the report, because one of the children results in each case; CME deemed an average wrote, “This new law the Communications of 4.1 of these contained important educa- Decency Act totally defys [sic] all that the tional information. Constitution was. Fight the system, take the Eddy Jansson and Matthew Skala, The Break- power back. ...” ing of Cyber Patrol ®4 (Mar. 11, 2000) Bennett Haselton, “Cyber Patrol Error Rate Jansson and Skala decrypted Cyber Patrol’s for First 1,000 .com Domains” (Peacefire, blacklist and found questionable blocking of Oct. 23, 2000) Peacefire, as well as a number of anonymizer and foreign-language translation services, Haselton tested Cyber Patrol’s average rate which the company blocked under all of its de- of error, using the same 1,000 dot-com do- fault categories. Blocked under every category mains as a sample that he used for an identical but “sex education” was the Church of the Sub- investigation of SurfWatch (see page 36). The Genius site, which parodies Christian churches CyberNOT list blocked 121 sites for portray- as well as corporate and consumer culture. als of “partial nudity,” “full nudity,” or “sexual acts.” Of these 121 sites, he eliminated 100 Also on the block list, for “intolerance,” that were “under construction,” and assessed were a personal home page on which the word the remaining 21. He considered 17 wrongly “voodoo” appeared (in a mention of voodoo- blocked, including a-actionhomeinspection. cycles.com) and the Web archives of Declan com; a-1bonded.com (a locksmith’s site); McCullagh’s Justice on Campus Project, a-1janitorial.com; a-1radiatorservice.com; which worked “to preserve free expression and and a-attorney-virginia.com. He deemed due process at universities.” Blocked in the four sites appropriately blocked under Cyber “satanic/cults” category were Webdevils.com Patrol’s definition of sexually explicit content, (a site of multimedia Net-art projects) and for an error rate of 81%. Haselton wrote that Mega’s Metal Asylum, a page devoted to heavy Cyber Patrol’s actual error rate was anywhere metal music; the latter site was also branded between 65-95%, but was unlikely to be “less “militant/extremist.” Also blocked as “mili- than 60% across all domains,” and as with tant/extremist,” as well as “violence/profanity” Bess, that the results may have been skewed in and “questionable/illegal & gambling,” were a Cyber Patrol’s favor owing to the test’s focus portion of the Nuclear Control Institute site; on dot-com domains, which “are more likely a personal page dedicated, in part, to rais- to contain commercial pornography than, say, ing awareness of neo-Nazi activity; multiple .org domains.” editorials opposing nuclear arms from Wash- Brennan Center for Justice 19 Bennett Haselton & Jamie McCarthy, “Blind plicit” content: Amnesty International Israel; Ballots: Web Sites of U.S. Political Candidates the Canadian Labour Congress; the American Censored by Censorware” (Peacefire, Nov. 7, Kurdish Information Network, which tracks 2000) human rights violations against Kurds in Iran, Iraq, Syria, and Turkey; the Milarepa Fund, In this Election Day report, Peacefire re- a Tibetan interest group; Peace Magazine; vealed that Cyber Patrol, configured to block the Bonn International Center for Conver- “partial nudity,” “full nudity,” and “sexual sion, which promotes the transfer of human, acts,” blocked the Web sites of four Repub- industrial, and economic resources away from lican candidates, four Democrats, and one the defense sector; the Canada Asia Pacific Libertarian. The site of an additional Demo- Resource Network, whose stated mission “is cratic candidate, Lloyd Doggett, was blocked to promote regional solidarity among trade under Cyber Patrol’s “questionable/illegal/ unions and NGOs in the Asia Pacific” region; the Sisterhood Is Global Institute, an orga- Cyber Patrol, configured to nization opposing violations of the human block “partial nudity,” “full rights of women worldwide; the Metro Network for Social Justice; the Society nudity,” and “sexual acts,” for Peace, Unity, and Human Rights for blocked the Web sites of Sri Lanka; and the International Coptic Congress. four Republican candidates, “Digital Chaperones for Kids,” Consumer four Democrats, and one Reports (Mar. 2001) Libertarian. Consumer Reports found that Cyber Patrol failed to block 23% of the magazine’s cho- gambling” category. The day after Peacefire sen 86 “easily located Web sites that contain published these findings, ZDNet News re- sexually explicit content or violently graphic porter Lisa Bowman contacted Cyber Patrol’s images, or that promote drugs, tobacco, then-manufacturer, SurfControl. A company crime, or bigotry.” Yet it did block the home spokesperson directed Bowman to the Cy- page of Operation Rescue (which the authors berNOT search engine, which indicated that classified as objectionable on account of its none of the URLs was actually prohibited. graphic images of aborted fetuses). The filter But later the same day, after downloading also blocked such nonpornographic sites as Cyber Patrol’s most recent block list, Bow- Peacefire and Lesbian.org. man attempted to access each site, and found that the software did indeed bar her from the Kieren McCarthy, “Cyber Patrol Bans The candidate sites in question.50 Register,” The Register (Mar. 5, 2001); Drew Cullen, “Cyber Patrol Unblocks The Register,” “Amnesty Intercepted: Global Human Rights The Register (Mar. 9, 2001) Groups Blocked by Web Censoring Software” (Peacefire, Dec. 12, 2000) Days after the Consumer Reports article appeared, the British newspaper The Register “Amnesty Intercepted” reported the follow- received word from an employee of Citrix ing organizations (among others) blocked by Systems that he had been unable to access the Cyber Patrol in the category of “sexually ex- Register from his office computer, on which 50 Peacefire, “Inaccuracies in the ‘CyberNOT Search Engine,’” the company had installed Cyber Patrol. (n.d.); Lisa Bowman, “Filtering Programs Block Candidate SurfControl unblocked the site within days, Sites,” ZDNet News (Nov. 8, 2000). 20 Internet Filters: A Public Policy Report with the exception of a page containing the had selected as nonobscene, including the December 12, 2000, article that was the basis sex information sites Di Que Si; All About of the initial block: a piece by Register staff Sex; New Male Sexuality; and Internet Sex reporter John Leyden on Peacefire’s recently Radio. introduced filter-disabling program.51 • In the New York Times article “Library A SurfControl representative explained: Grapples with Internet Freedom” (Oct. “The Register published an article written by 15, 1998), Katie Hafner reported that Peacefire containing information on how to Cyber Patrol blocked searches for Georgia access inappropriate sites specifically blocked O’Keeffe and Vincent van Gogh, while by Cyber Patrol. Given [the] irresponsible allowing hits from searches for “toys” that nature of the article, apparently encourag- included sites selling sex toys. ing users to override Cyber Patrol’s filtering • Peacefire reported, in “BabelFish Blocked mechanism, we took the decision to block by Censorware” (Feb. 27, 2001), that Cyber The Register – upholding our first obliga- Patrol blocked the foreign-language Web tion to customers by preventing children page translation service featured on Alta- or pupils from being able to surf Web sites Vista in all 12 filtering categories. containing sexually explicit, racist or inflam- matory material.” Cullen responded that • In “Teen Health Sites Praised in Article, there was no “sexually explicit, racist,” or Blocked by Censorware” (Mar. 23, 2001), “inflammatory” material in the article, which Bennett Haselton reported that Cyber “merely describes peacefire.exe and provides Patrol blocked ZapHealth, a health educa- a link to the Peacefire.org Web site.”52 tion site containing articles of interest to a teenage audience. Miscellaneous reports • In “Censorware: How Well Does Inter- Cyber Sentinel net Filtering Software Protect Students?” Rather than maintaining and updating a list (Jan. 1998), Electronic School columnist of sites to be blocked, or designating forbid- Lars Kongshem reported that Cyber Patrol den categories, Cyber Sentinel scans each blocked the “Educator’s Home Page for requested page for key words and phrases in Tobacco Use Prevention,” part of a site its various databases, or “libraries.” In 2001, maintained by Maryland’s Department of for example, its “child predator library” Health and Mental Hygiene. contained such phrases and “do you have a pic” and “can I call you.” Promotional text • In his expert witness report for the defen- on Cyber Sentinel’s Web site claims it is “the dants in the case of Mainstream Loudoun v. most advanced and flexible” Internet filtering Board of Trustees of the Loudoun County Li- product on the market.53 brary (July 14, 1998), David Burt reported that his comparative testing of Cyber Patrol, Center for Media Education, Youth Access to I-Gear, SurfWatch, and X-Stop revealed Alcohol and Tobacco Web Marketing: The Filter- that Cyber Patrol blocked 40% of sites Burt ing and Rating Debate (Oct. 1999) 51 J ohn Leyden, “Porn-filter Disabler Unleashed,” The Register The CME found Cyber Sentinel ineffec- (Dec. 19, 2000). 52 e SurfControl representative also wrote: “We should be Th 53 “ Cyber Sentinel Filtering Network,” www.cyber-sentinel. grateful if The Register would adopt a policy of allowing net (visited 3/7/06). This site is operated by Software4Par- companies, such as ourselves, the opportunity to respond in ents.com. Another product, “Cyber Sentinel Network,” is full before going to press.” “Astonishing,” Cullen commented. made by Security Software Systems and seems to use similar “Cyber Patrol blocked The Register without informing us, or technology, adjusted for office rather than home use, www. giving us a chance to respond in full, or at all.” securitysoft.com/default.asp?pageid=49 (visited 3/7/06). Brennan Center for Justice 21 tive in screening out promotions for alcohol Children in Cyberspace, an appendix of which and tobacco use. It blocked only 11% of is titled “Porn on the Net.” the promotional sites selected by the testers, allowing users to access an average of 39 of CYBERsitter the 44 pages, and blocked just 3% of the Before 1999, CYBERsitter, in addition to pages resulting from searches for alcohol- and blocking entire sites and searches for terms on tobacco-related promotional material. its block list, would excise terms it deemed objectionable and leave blank spaces where Bennett Haselton, “Sites Blocked by Cyber they would otherwise appear. This procedure Sentinel” (Peacefire, Aug. 2, 2000) led to some early notoriety for the product, Having conducted “about an hour of ad- as when it deleted the word “homosexual” hoc experimentation,” Haselton found that from the sentence, “The Catholic Church Cyber Sentinel blocked CNN because, as opposes homosexual marriage” – and left Web system log files revealed, the word “erotic” users reading “The Catholic Church opposes appeared on the front page (in the title of marriage.”54 an article, “Naples Museum Exposes Public In 1999, CYBERsitter, manufactured by to Ancient Erotica”). Also blocked were: a Solid Oak Software, modified its system and result page for a search of the word “censor- established seven default settings, including ship” on Wired magazine’s site (one of the “PICS rating adult topics,” which covered “all results contained the word “porn” in the title, topics not suitable for children under the age of “Feds Try Odd Anti-Porn Approach”); result 13,” “sites promoting the gay and lesbian life- pages for searches of the term “COPA” on style,” and “sites advocating illegal/radical activi- Wired and other news sites, also on account of ties.” Its total list of blocking categories grew to article titles containing the word “porn” (for 22. Users could enable or disable any category. instance, “Appeals Court Rules Against Net Porn Law”); and a portion of the Web site of By 2005, CYBERsitter had 32 content the Ontario Center for Religious Tolerance, categories, including “cults,” “gambling,” and containing an essay on collisions between sci- “file sharing.” Its default setting blocked “sex,” ence and religion throughout history. “violence,” “drugs,” and “hate,” as well as all image searches. PC Magazine called it “the Cyber Sentinel also blocked sites associ- strongest filtering we’ve seen. … CYBERsitter ated with both sides of the civil liberties and errs on the conservative side.” It also blocked Internet censorship debates: an ACLU press “bad words” in email and instant messages.55 release, “Calls for Arrest of Openly Gay GOP Convention Speaker Reveal Danger of Sod- Brock Meeks & Declan McCullagh, “Jack- omy Laws Nationwide”; the American Family ing in From the ‘Keys to the Kingdom’ Port,” Association, because of the word “porn” CyberWire Dispatch (July 3, 1996) (“the current administration and the Justice Meeks and McCullagh reported that CY- Department have been good to the porn BERsitter blocked a newsgroup devoted to industry”); on account of the word “cum,” the gay issues (alt.politics.homosexual), the Queer biographies of COPA Commission members Resources Directory, and the home page Stephen Balkam and Donna Rice Hughes of the National Organization for Women. (both graduated magna cum laude); the COPA CYBERsitter’s prohibited words included Commission’s list of research papers, because the word “porn” appeared in the title of one 54 Peacefire, “CYBERsitter Examined” (2000). report; and the home page for Donna Rice 55 “ CYBERsitter 9.0,” PC Magazine (Aug. 3, 2004), www.pcmag.com/ar- ticle2/0,1759,1618830,00.asp (visited 2/1/06); see also “CYBERsitter® Hughes’s book, Kids Online: Protecting Your – For a Family Friendly Internet,” CYBERsitter.com (visited 2/10/06). 22 Internet Filters: A Public Policy Report “gay, queer, bisexual” combined with “male, mail in return demanding that I cease writing men, boy, group, rights, community, activi- to them and calling my mail ‘harassment’ ties…” and “gay, queer, homosexual, lesbian, —with a copy to the postmaster at my ISP.” bisexual” combined with “society, culture.” Karen Schneider, A Practical Guide to Internet According to the report, Brian Milburn, Filters (1997) president of Solid Oak Software, responded: Schneider’s Internet Filter Assessment “We have not and will not bow to pressure Project reported that unlike other filtering from any organization that disagrees with our products, CYBERsitter does not permit its philosophy. ... We don’t simply block pornog- keyword-blocking feature to be disabled. raphy. That’s not the intention of our product. Regarding CYBERsitter’s claim that it “looks The majority of our customers are strong at how the word or phrase is used in context,” family-oriented people with traditional family Schneider quoted one TIFAP tester: “Noth- values. ... I wouldn’t even care to debate if gay ing could be further from the truth.” The and lesbian issues are suitable for teenagers. filter deleted the word “queer,” for example, … We filter anything that has to do with sex. from Robert Frost’s “Stopping by Woods on a Sexual orientation [is about sex] by virtue of Snowy Evening” (“My little horse must think the fact that it has sex in the name.” it queer / To stop without a farmhouse near”). Bennett Haselton, “CYBERsitter: Where Do CYBERsitter did not block sites containing We Not Want You To Go Today?”, Ethical instructions for growing marijuana, but did Spectacle (Nov. 5-Dec. 11, 1996) block a news item on the legislation surround- ing it. Haselton reported that among CYBER- sitter’s blocked domains were, in addition Marie José Klaver, “What Does CYBERsitter to Peacefire itself, the online communities Block?” (June 23, 1998) Echo Communications and Whole Earth In June 1998, Marie-José Klaver decrypted ’Lectronic Link; the Web site of Community and published CYBERsitter’s full list of ConneXion, which manufactured an anony- blocked words, strings, sites, and domains. mous-surfing program; and the home page Among the domains on the block list were of the National Organization for Women. servers of the University of Chicago, the CYBERsitter also barred any Yahoo search for University of Virginia’s Information Technol- the term “gay rights.” ogy & Communication Division, Georgia Ethical Spectacle press release, “CYBERsitter State University, the University of Michigan’s Blocks The Ethical Spectacle” (Jan. 19, 1997) engineering department, and Rutgers Univer- sity; several large Dutch domains, including In early 1997, CYBERsitter blocked the euronet.nl, huizen.dds.nl, and worldaccess.nl; Ethical Spectacle, an online magazine “examin- and the phrases “bennetthaselton,” “peacefire,” ing the intersection of ethics, law and politics and “dontbuyCYBERsitter.” in our society,” after editor Jonathan Wallace added a link to a site titled “Don’t Buy CY- Christopher Hunter, Filtering the Future (July BERsitter,” which directed users to Peacefire’s 1999) report “CYBERsitter: Where Do We Not Though Christopher Hunter’s study (see Want You to Go Today?” Wallace wrote to page 17) concluded that CYBERsitter was the Milburn and Solid Oak technical support “de- most reliable filter at screening out “objection- manding an explanation. I pointed out that able” sites (it blocked 25, or 69.4%, of such The Spectacle does not fit any of their pub- sites in his sample), he also noted that the lished criteria for blocking a site. I received Brennan Center for Justice 23 software performed well below the 90-95% .com. While performing better than most rate of accuracy boasted by the manufacturer. other filters in its response to searches for CYBERsitter fared worst in its treatment of promotional content – CYBERsitter prohib- “nonobjectionable” material, blocking 24, or ited searches for “beer,” “cigarettes,” “cigars,” 14.6%, of the sites to which Hunter assigned and “liquor” – it subsequently blocked just RSAC ratings no higher than one. Among 3% of the result pages (from the allowed these were Sharktagger, a site promoting searches) that the CME testers attempted to responsible shark fishing and conservation; view. CYBERsitter also blocked 13% of the a listing of local events posted on Yahoo; CME’s chosen educational and public health RiotGrrl; Planned Parenthood; Stop Prisoner sites, and prohibited testers from conduct- Rape; the National Organization for Women; ing searches for “alcohol,” “alcoholism,” “fetal the feminist performance-art and activist alcohol syndrome,” “tobacco,” and “tobacco troupe Guerrilla Girls; the Church of Scien- settlement.” tology; The Body, an informational site on Peacefire, “CYBERsitter Examined” (2000) AIDS and HIV; Williams College’s informa- tion page on safe sex; the Coalition for Posi- The original report on this study described tive Sexuality, SIECUS, and Pro-Life America. CYBERsitter’s blocking of numerous non- profit sites, including the Penal Lexicon, a Cybersitter left Web users British project documenting prison conditions reading “the Catholic worldwide; the Department of Astronomy at Smith College; the Computer Animation Church opposes marriage.” Laboratory at the California Institute of the Arts; and the College of Humanities & Social CYBERsitter proved particularly likely to Sciences at Carnegie Mellon University. The deny access to nonpornographic sites relating filter’s by-then well-known propensity for to homosexuality, blocking the QWorld con- overblocking resulted from its keyword-based tents page; the gay communities Planet Out, technology, combined with the manufacturer’s PrideNet, and the Queer Zone; A Different decision to block sites like the National Or- Light Bookstore, which specializes in gay and ganization for Women in order to appeal to lesbian literature; Gay Wired Presents Wildcat a rightwing constituency. The current text of Press; and Queer Living’s “Promoting with “CYBERsitter Examined” recounts the history Pride” page. (These sites, while not falling of CYBERsitter’s disputes with its critics, and under RSAC’s definition of unacceptability, links to lists of previously blocked sites. do fall within CYBERsitter’s default filter- ing category of “sites promoting the gay and Bennett Haselton, “Amnesty Intercepted: lesbian lifestyle.”) Global Human Rights Groups Blocked by Web Censoring Software” (Peacefire, Center for Media Education, Youth Access Dec. 12, 2000) to Alcohol and Tobacco Web Marketing (Oct. 1999) CYBERsitter blocked a number of pages on the Amnesty International site because of its The CME charged CYBERsitter with keyword filtering mechanism. A news item under- and overinclusive filtering of alcohol- containing the sentence, “Reports of shoot- and tobacco-related material, as it blocked ings in Irian Jaya bring to at least 21 the num- only 19% of the promotional sites in the test ber of people in Indonesia and East Timor sample – leaving unblocked beer sites such killed or wounded,” was prohibited for its as heineken.com, and sites on which tobacco “sexually explicit” content. Peacefire’s review products were sold, such as lylessmokeshop 24 Internet Filters: A Public Policy Report of the system log revealed that CYBERsitter member Pamela Arn – presumably because had blocked the site after detecting the words CYBERsitter detected its blacklisted phrase “least 21.” The filter blocked another human “pamela.html” in the URL and confused rights page, which noted that the United Na- the site with one devoted to Pamela Ander- tions Convention on the Rights of the Child son; and part of iEmily.com, including the “defines all individuals below the age of 18 Terms of Service page, on which the words years as children,” for the words “age of 18.” “sexually oriented” appeared. (One of the terms of service is that users “will not use “Digital Chaperones for Kids,” Consumer [iEmily’s] message boards or chat rooms Reports (Mar. 2001) to post any material which is ... sexually While failing to block 22% of sites that oriented.”) Consumer Reports deemed objectionable • In a Censorware Project post, “Columnist because of “sexually explicit content or Opines Against Censorware, Gets Column violently graphic images” or promotion of Blocked” (Mar. 29, 2001), Bennett Hasel- “drugs, tobacco, crime, or bigotry,” CYBER- ton reported that CYBERsitter blocked sitter blocked nearly one in five of the sites “Web Filters Backfire on their Fans,” a Chi- the authors considered inoffensive, including cago Tribune article that criticized filtering Lesbian.org, the Citizens’ Committee for the software, apparently because the software Right to Keep and Bear Arms, and the South- detected the words “porno,” “Internet ern Poverty Law Center. porn,” and “Peacefire” in the article. Miscellaneous reports • A short article by Greg Lindsay in Time • In a review of “Filtering Utilities” (Apr. 8, Digital (Aug. 8, 1997), referenced by Peace- 1997), PC Magazine noted that CYBERsit- fire’s “Blocking Software FAQ” reported ter blocked an engineering site with “Bour- that CYBERsitter blocked a Time magazine bonStreet” in its URL. article critical of the filter. • According to the Digital Freedom Net- work’s “Winners of the Foil the Filter FamilyClick FamilyClick, whose spokesperson was anti- Contest” (Sept. 28, 2000), CYBERsitter pornography activist Donna Rice Hughes, blocked House Majority Leader Richard allowed users to choose from five configura- “Dick” Armey’s official Web site upon tions. Its least restrictive “Full FamilyClick detecting the word “dick,” and Focus on the Access” setting, recommended for ages 18+, Family’s Pure Intimacy page, which protests blocked sites falling into any of seven cat- pornogaphy and is geared toward individu- egories, including “crime,” “gambling,” and als “struggling with sexual temptations.” “chat.” Its “Teen Access” setting, for ages • In “Teen Health Sites Praised in Article, 15-17, blocked the previous seven categories Blocked by Censorware” (Mar. 23, 2001), plus “personals,” “illegal drug promotion,” Peacefire’s Bennett Haselton reported that “chat/message boards,” and “non-Family- CYBERsitter barred part or all of three Click email services.” “Preteen Access,” for out of the four sites discussed in a recent ages 12-14, barred four additional catego- New York Times article on health education ries, including “advanced sex education” and resources for teenagers: ZapHealth; various “weapons.” “Kids Access,” geared toward ages pages on Kidshealth.org, including its anti- 8-11, blocked “basic sex education.” Finally, smoking page, a page of advice on travel the “Children’s Playroom,” for ages seven safety, and a profile of KidsHealth staff and under, was described as “100% safe. It Brennan Center for Justice 25 contains activities, games and content that site called “Countering Terrorism,” regarding have been pre-selected and pre-approved by the slaying of 11 Israeli athletes at the 1972 FamilyClick.” Munich Olympics; and a guide to “Eating Right” for chronic obstructive pulmonary The filter’s Web site in 2005 stated that disease patients. “FamilyClick has ceased operating its filtering service until further notice.”56 Online Policy Group, “Online Oddities and Atrocities Museum” (n.d.) Bennett Haselton, “Sites blocked by Family- Click” (Peacefire, Aug. 1, 2000) The Online Policy Group maintained, as part of its “Online Oddities and Atrocities Haselton conducted “about an hour’s worth Museum,” a list of sites at different times of ad-hoc testing” of FamilyClick on its mistakenly blocked by FamilyClick. These least restrictive “18+” setting and found that included the Christian Coalition (which is among the sites blocked were: a report from headed by FamilyClick founder Tim Rob- the U.S. embassy in Beijing on the AIDS epi- ertson’s father, Pat Robertson), The Oprah demic in China; a research study on gambling Winfrey Show, which was blocked, in the in Washington State; a Spanish-language glos- midst of a product demonstration by Donna sary of AIDS terms; the home page of Camp Rice Hughes during her appearance on that Sussex, which organizes summer programs for program; and the FamilyClick site itself. children of low-income households; Psyart, an online journal published by the University of Florida’s Institute for the Psychological Study I-Gear of the Arts; an essay titled “Triangles and I-Gear, manufactured by the Symantec Cor- Tribulations: The Politics of Nazi Symbols,” poration, as of 2001 operated through a com- on a Holocaust Studies site; a Christian Re- bination of predefined URL databases and search Journal article condemning homosexu- “Dynamic Document Review.” As it described ality; and the genealogical page for one Alice the process, I-Gear divided its site database Ficken – perhaps, Haselton observed, because into 22 categories. If a URL was not in any “Ficken” is the infinitive form of “fuck” in of the databases, I-Gear scanned the page German. for trigger words from its “DDR Dictionar- ies.” Each matching word on a site received a I-Gear barred searches on numerical score; if the total score for the page exceeded 50 (the default maximum score, eating disorders, AIDS, which could be adjusted to anywhere between and child labor. 1-200), the site was blocked. Other blocked sites included an inventory By 2005, Symantec had reconfigured I-Gear of state sodomy laws on the Web site of the to be a component of larger products such as ACLU; a genealogical index of individuals “Symantec Web Security,” which combines bearing the name “Mumma”; a background filtering out “inappropriate” content with report on pornography by the Minnesota anti-virus and other noncensorship-based Family Council; a page on the Ontario Center protections.57 We were unable to find any de- for Religious Tolerance site that tracked anti- scription of the filtering method used, beyond Wiccan content on Christian Web sites; an the assurance that Symantec Web Security essay on the Federation of American Scientists “ensures maximum protection by combining 56 Family Click: Your Guide to the Web,” www.familyclick.com 57 “Symantec Web Security,” enterprisesecurity.symantec.com/prod- (visited 3/13/05). ucts/products.cfm?ProductID=60 (visited 3/14/05). 26 Internet Filters: A Public Policy Report list-based techniques with heuristic, context- were “obvious errors” and 10, “marginal er- sensitive analysis tools.”58 rors” (blocking of sites with moderately adult content but not falling within I-Gear’s defini- Karen Schneider, A Practical Guide to Internet tion of “sexual acts”). Among the “obvious” Filters (1997) wrongful blocks were sites containing refer- Schneider suggested that I-Gear’s state-of- ences to or information on homosexuality, the-art-sounding Dynamic Document Review such as the personal home page of Carnegie basically amounted to keyword blocking. Mellon Robotics Institute programmer Duane For this reason, TIFAP tested I-Gear under T. Williams and an anti-gay pamphlet, posted its least restrictive “adult” setting with DDR on the Web site of the Gay and Lesbian Alli- disabled, thus using only the list of pro- ance at the Georgia Institute of Technology. scribed URLs. It found that, even with this Also blocked were sites with anti-censorship configuration, I-Gear blocked the entire Gay content; astronomy, cartography, and art and Lesbian directory of Yahoo, as well as museum sites; and an essay on “Indecency on pages containing the words “cockfighting” and the Internet: Lessons from the Art World,” by “pussycat.” Julie Van Camp, a philosophy professor at the University of California. Anemona Hartocollis, “Board Blocks Student Access to Web Sites: Computer Filter Hobbles Other sites blocked for reasons unknown Internet Research Work,” New York Times included “Semi-Automatic Morph Between (Nov. 10, 1999) Two Supermodels,” an animation sequence written by an MIT student in which images The New York Times reported that I-Gear of two models’ faces morph into each other; a barred students at Benjamin Cardozo High diagram of a milk pasteurization system; a site School in New York City from conducting containing Book X, in Latin, of the Confes- searches on such topics as breast cancer, eating sions of St. Augustine – possibly because of the disorders, AIDS, and child labor. Though Sy- common appearance of the word “cum”;59 two mantec senior product manager Bernard May pages on the Wheaton College server contain- responded to the news by insisting that I-Gear ing sections of The Decline and Fall of the Ro- demonstrated “absolutely no preference of one man Empire; and lecture notes from a philoso- group or another,” the article also noted that phy course at the University of Notre Dame. I-Gear blocked the pro-choice groups Planned Parenthood and Alan Guttmacher Institute, Peacefire, “I-Gear Examined” (2000) but not Right to Life and Operation Rescue. In tests of I-Gear throughout the first half Students were also unable to access portions of 2000, Peacefire found more sites blocked of an electronic text of The Grapes of Wrath for questionable reasons, including the full – specifically, “a passage in which a woman text of Jane Eyre; the federal district court’s lets a starving man suckle at her breast.” ruling on the Communications Decency Act Peacefire, “Analysis of First 50 URLs Blocked in ACLU v. Reno; transcripts of testimony by I-Gear in the .edu Domain” (Mar. 2000) from ACLU v. Reno; “Readings on Computer Communications and Freedom of Expres- Peacefire evaluated the first 50 dot-edu sites sion,” a supplementary reading list for a blocked in the “sexual acts” category accord- course on Internet ethics at MIT; and the free ing to a February 2000 I-Gear block list. Of speech page of the Center for Democracy and the 50 blocks, Peacefire determined that 27 Technology. 58 “ Symantec Web Security,” eval.veritas.com/mktginfo/enterprise/fact_ sheets/ent-factsheet_web_security_3.0_05-2005.en-us.pdf (visited 59 Chris Oakes, “Censorware Exposed Again,” Wired News 3/2/06). (Mar. 9, 2000) Brennan Center for Justice 27 I-Gear also barred a United Nations report, the Born-Again Virgins site; the Fine Art “HIV/AIDS: The Global Epidemic”; the Nude Webring; and the home pages of four Albert Kennedy Trust, which works on behalf fine art photography galleries. of homeless gay teenagers; the Anti-Violence Project, which opposes anti-gay violence; the • Peacefire’s “Amnesty Intercepted” (Dec. 12, International Gay and Lesbian Human Rights 2000) reported that I-Gear blocked the of- Committee; the Human Rights Campaign; ficial site of the 1999 International Confer- the Harvard Gay and Lesbian Caucus; two ence Combating Child Pornography on the pages of the National Organization for Internet. Women site, one providing information on • In the March 23, 2001 report “Teen Health gay rights, the other a press release on the Sites Praised in Article, Blocked by Cen- legal status of gay marriage in Hawaii; a state- sorware,” Bennett Haselton reported that ment on equal rights for homosexuals and I-Gear’s Dynamic Document Review led women in the workplace from the Industrial to the partial blocking of three sites lauded Workers of the World; a portion of GLAAD’s in a recent New York Times article describ- site containing information for prospective ing health education sites for teens: iEmily; volunteers; “The Homosexual Movement: A KidsHealth; and ZapHealth. Response,” a statement by a group of Jew- ish and Christian theologians, ethicists, and Internet Guard Dog scholars; two Web sites relating to the Chris- Internet Guard Dog, manufactured by tian Coalition – the organization’s legal arm, McAfee, claimed in 2001 that it had “a the American Center for Law and Justice, and comprehensive objectionable content the Pat Robertson-owned Christian Broad- database” that prevented “messages deemed casting Network; and the home page of the inappropriate … from reaching your British Conservative Party. child.” “Offensive words” as well as sites Other blocked sites included a Cato Insti- were blocked. A June 9, 2000 review noted tute policy paper titled “Feminist Jurispru- that Guard Dog allowed the user to filter dence: Equal Rights or Neo-Paternalism?”; by category (e.g., drugs, gambling, the a gender studies page on Carnegie Mellon occult) from levels 0 through 4, and that University’s English server; Planned Parent- “when a line contains a disallowed word, hood; CyberNOTHING’s critical commen- Guard Dog replaces the entire line with as- tary on a 1995 Time magazine cover story terisks.”60 By 2005, McAfee was no longer about pornography online; and the article marketing Internet Guard Dog, but instead “PETA and a Pornographic Culture,” which offered “McAfee Parental Controls”; we protested the use of nude models in recent could find no information about blocking PETA advertising campaigns, but contained categories.61 no nude images. Peacefire does not indicate which I-Gear categories were enabled when 60 Review of Internet Guard Dog, ZdNet (June 9, 2000), URL the various sites were blocked. no longer available. A March search found a ZdNet page noting that Guard Dog 3.0 “is not yet available from any of Miscellaneous reports our online merchants.” “Guard Dog 3.0 Win 9X,” shopper- zdnet.com.com/Guard_Dog_3_0_Win9X/4027-3666_15- 1587666.html?tag=search&part=&subj= (visited 3/10/06). • In his July 1998 expert witness report for 61 s.mcafee.com/root/package.asp?pkgid=146&WWW_ u the defendants in Mainstream Loudoun v. URL=www.mcafee.com/myapps/pc/default.asp (visited Board of Trustees of Loudoun County Library, 3/13/05). By 2006, this URL led to a different site, advertis- ing McAfee Privacy Service,” us.mcafee.com/root/package. David Burt reported that I-Gear blocked asp?pkgid=146&WWW_URL=www.mcafee.com/myapps/ pc/default.asp (visited 3/10/06). 28 Internet Filters: A Public Policy Report “Digital Chaperones for Kids,” Consumer tions of objectionable and nonobjectionable Reports (Mar. 2001) sites (see page 17), that Net Nanny’s major failing lay in its underinclusive blocking. The Guard Dog failed to block 30% of “easily software “performed horrendously,” he wrote, located Web sites that contain sexually explicit “blocking a measly 17% of objectionable content or violently graphic images, or that content” (it failed to block 30 sites, including promote drugs, tobacco, crime, or bigotry.” It www.xxxhardcore.com and www.ultravixen. did block nearly 20% of sites that the testers com). But it also blocked the fewest “nonob- deemed politically controversial but not por- jectionable sites” (3%). That 3% consisted of nographic or violent, including the National the Queer Resources Directory; the official Institute on Drug Abuse and SEX, Etc., the home page of the White House; the Web site Rutgers University educational site written by of Northwestern University professor and Ho- and for teens. locaust revisionist Arthur Butz; the Adelaide Institute, another revisionist history site; an Net Nanny online casino; and the Coalition for Positive Net Nanny advertises itself as “the Web’s origi- Sexuality. nal Internet filter,” first launched in January 1994.62 It is a freestanding software product Center for Media Education, Youth Access that blocks based on “known inappropriate to Alcohol and Tobacco Web Marketing key words and phrases” and, as of 2005, on a (Oct. 1999) monthly updated block list. Customers can Net Nanny allowed every search that the customize their block lists, filter settings, and CME attempted, for both promotional and keyword lists.63 As of 2006, Net Nanny had educational alcohol- and tobacco-related five blocking categories: “sexual explicitness,” sites. It blocked just 2% of the promotional “hate,” “violence,” “crime,” and “drugs.” sites in the test sample. Though initially un- While generally commended for its will- able to access the Cuervo Tequila site, CME ingness to disclose its lists, Net Nanny has researchers easily viewed it by entering the nonetheless fared poorly in studies, with high page’s numerical IP address instead of its rates of both over- and underblocking. alphabetical one. Thus, “it would appear that Net Nanny does not regularly take IP ad- Karen Schneider, A Practical Guide to Internet dresses into consideration when compiling Filters (1997) its blacklist” – making it easy to circumvent. Schneider reported that Net Nanny blocked What Net Nanny did block was health.org, a long-obsolete URL containing artistic apparently because its front page title, “Drug erotica, which was part of an early version AbuseXXXXXXXXXX,” was detected by the of Yahoo, but did not block www.creampie. product’s keyword-blocking feature. com, a sexually explicit site that had been in Peacefire, “Net Nanny Examined” (2000) existence for six months. Among the newsgroups that Peacefire found Christopher Hunter, Filtering the Future inappropriately blocked by Net Nanny were (July 1999) bit.listserv.aidsnews; sci.med.aids; and alt. Hunter concluded, based on his designa- feminism. Other blocked sites included the Banned Books page at Carnegie Mellon Uni- 62 “ Net Nanny is CIPA-Compliant and More,” www.netnanny. com/p/page?sb=cipa (visited 2/22/06). versity and Femina.com, a “comprehensive, 63 “ Find Out More About Net Nanny,” www.netnanny.com/ searchable directory of links to female friendly page?sb=detailed, and www.netnanny.com/NetNanny5/as- sites and information.” Peacefire observed that sets/docs/nn5/userguide.pdf (visited 2/22/06). Brennan Center for Justice 29 “while the Banned Books page and Femina. Net Shepherd com are blocked because the URLs exist as In October 1997, the AltaVista search engine entries on Net Nanny’s blocked site list, more and the organization Net Shepherd launched Web sites are blocked because they contain a “Family Search” function to screen the keywords which activate Net Nanny’s word results of AltaVista searches according to filter.” NetShepherd’s database of site ratings. Net “Digital Chaperones for Kids,” Consumer Shepherd claimed to have rated more than Reports (Mar. 2001) 300,000 sites based on “quality” and “matu- rity,” relying on “demographically appropriate Consumer Reports reinforced Hunter’s and Internet users’” judgments of what would be the CME’’s conclusions, reporting that Net “superfluous and/or objectionable to the aver- Nanny failed to block 52% of 86 “easily age family.”64 located Web sites” selected by the magazine that “contain sexually explicit content or vio- By 2005, Net Shepherd seemed to be out of lently graphic images, or that promote drugs, business. The original URLs for the product tobacco, crime, or bigotry.” The filter did, led to alternative sites, and a “Net Shepherd however, block Rutgers University’s teen-ori- World Opinion Rating” site describing the ented SEX, Etc. software had not been updated since Novem- ber 1997.65 Miscellaneous reports Electronic Privacy Information Center • According to Brock Meeks and Declan Mc- (EPIC), Faulty Filters: How Content Filters Cullagh’s “Jacking in From the ‘Keys to the Block Access to Kid-Friendly Information on the Kingdom’ Port” (July 3, 1996), Net Nanny Internet (1997) blocked every mailing list originating at cs.coloradu.edu, the computer science divi- In November 1997, EPIC performed the sion of the University of Colorado, as well same 100 searches on standard AltaVista and as unspecified “feminist newsgroups.” the “Family Search” version. Its sample of search terms included 25 schools, 25 charita- • In “Adam and Eve Get Caught in ’Net ble and political organizations, 25 educational Filter” (Feb. 5, 1998), the Wichita Eagle and cultural organizations, and 25 “miscella- reported that students at Wichita’s Friends neous concepts and entities” that could be University, where Net Nanny had been of research interest to children, for instance installed at public computer stations, “astronomy,” “Bill of Rights,” “Teen were barred from accessing pages con- Pregnancy,” and “Thomas Edison.” taining educational material on sexually transmitted diseases, prostitution, and The first search term on EPIC’s list was Adam and Eve. “Arbor Heights Elementary.” This primary school’s site contained, among other features, • Net Nanny was cited in the Digital Free- an online version of Cool Writers Magazine, dom Network’s “Winners of the Foil the a literary periodical for ages 7-12. The search Filter Contest” (Sept. 28, 2000) for block- on unfiltered AltaVista resulted in 824 sites ing House Majority Leader Richard “Dick” mentioning the school, while the same search Armey’s official Web site upon detecting the through Net Shepherd returned only three. word “dick.” It also precluded a high school student in Australia from accessing sites on 64 Net Shepherd press release (Dec. 1997). the genetics of cucumbers once Net Nanny 65 “ Net Shepherd World Opinion Rating Service,” www. detected the word “cum” in “cucumbers.” research.att.com/projects/tech4kids/Net_Shepherd_World_ Opinion_Rating_Service.html (visited 3/13/05; 2/26/06). 30 Internet Filters: A Public Policy Report Thus, EPIC determined that 99.6% function allowing users to unblock mistakenly of search results were filtered out. In blocked sites.66 subsequent searches for elementary, middle, “Digital Chaperones for Kids,” Consumer and high schools, EPIC concluded that Reports (Mar. 2001) Net Shepherd blocked between 86-99% of relevant material. Consumer Reports found that the NIS Fam- ily Edition left unblocked 20% of the sites the EPIC found similar filtering of informa- magazine deemed objectionable, while block- tion about charitable and political organiza- ing such organizations as the Citizens’ Com- tions, ranging from 89-99.9%. Among the mittee for the Right to Keep and Bear Arms most heavily filtered search results were those and the National Institute on Drug Abuse. for “American Cancer Society” (which had 38,762 relevant sites on an unfiltered search but only six with Net Shepherd), “United SafeServer As of 2005, SafeServer relied on “Intelligent Jewish Appeal” (for which Net Shepherd Content Recognition Technology,” which it allowed only one of the 3,024 sites that were described as “a leading-edge technology based otherwise reported as relevant); and “United on artificial intelligence and pattern recogni- Way” (for which Net Shepherd allowed 23 tion ... trained to detect English-language out of 54,300 responsive sites). Net Shepherd pornography” and to screen requested Web also filtered between 91-99.9% of relevant pages in real time. Hence, its advertisement: educational, artistic, and cultural institutions. “No lists, no subscriptions. Just fast, reliable On a search for “National Aquarium,” it al- filtering.” It had seven categories of blockable lowed 63 of the 2,134 sites otherwise reported content: “hate,” “pornography,” “gambling,” by AltaVista. Similarly, it blocked 99.5% of “weapons,” “drugs,” “job search,” and “stock sites responsive to the search term “photosyn- trading.”67 thesis,” 99.9% for “astronomy,” and 99.9% for “Wolfgang Amadeus Mozart.” Bennett Haselton, “SafeServer Error Rate for First 1,000 .com Domains” (Peacefire, The authors note some limitations in this Oct. 23, 2000) study, including the fact that the blocking percentages could be lower than reported On a SafeServer proxy in use at a high because even unfiltered AltaVista would not school where a student had volunteered to have provided access to all the responsive sites. assist with the research, Peacefire attempted Nevertheless, the obvious inference from this to access a selection of commercial domains study is that Net Shepherd was filtering out all used earlier in an identical test of SurfWatch, of the Internet except for a relative handful of and also in concurrent tests of Cyber Patrol approved, or “whitelisted,” sites. and AOL Parental Controls. SafeServer was configured to bar the categories of “drugs,” Norton Internet Security Family “gambling,” “hate,” “pornography,” and Edition (NIS) “weapons.” The filter blocked 44 pages in the The Norton Internet Security Family Edition 1,000-site sample; 15 of those 44 were “under is manufactured by Symantec, which also construction.” Haselton determined that of produces I-Gear (now embedded in Symantec 66 “ TopTen Reviews,” internet-filter-review.toptenreviews.com/ Web Security). The methodology and block- (visited 7/21/05); “Norton Internet Security,” www.symantec. com/home_homeoffice/products/internet_security/nis- ing categories are the same as I-Gear’s. As 30mac/features.html (visited 2/27/06). of 2005, the program had 31 categories of 67 “ Foolproof SafeServer,” smartstuff.com/safeserver/fepsafeserv. disapproved content. There was no override html; “Frequently Asked Questions,” smartstuff.com/safe- server/fpserverfaq.html (both visited 7/21/05). Brennan Center for Justice 31 the remaining 29 sites, 10 were inappropriate- opposition to Web censorship and filtering, ly blocked: a-1autowrecking.com; a-1coffee. including: the Electronic Frontier Founda- com; a-1security.com; a-1upgrades.com; a-2- tion’s Internet Censorship and Regulation r.com; a-abacomputers.com; a-artisticimages. archive; a list of free speech links on the Web com; a-baby.com; a-build.com; and a-c-r.com. site of the American Communication Associa- As with other filters tested, Haselton acknowl- tion; “The X-Stop Files” and “The Mind of a edged that the resulting error rate of 34% Censor,” two articles in The Ethical Spectacle; was not entirely reliable because of the small a CNet news article on guidelines drafted by sample under review, and that SafeServer the American Library Association for coun- could have an error rate as low as 15%; but tering campaigns for mandatory filtering ; that the rate of error among .org sites would the Wisconsin Civil Liberties Union and the presumably be higher, since pornographic National Coalition Against Censorship; and sites were more prevalent among commercial a Scientific American article on “Turf Wars in domains. Cyberspace.” SafeSurf also blocked the online edition of Free Inquiry, a publication of the SafeSurf Council for Secular Humanism; a United SafeSurf operates a voluntary self-rating sys- Nations paper on “HIV/AIDS: The Global tem. Authors of Web pages can evaluate their Epidemic”; and the full texts of The Odyssey sites according to 10 content categories, in- and The Iliad, both of which appeared on cluding “profanity,” “nudity,” “glorifying drug the University of Oregon server. Since it is use,” and “other adult themes.” In addition, unlikely that any of these sites self-rated, the each page is assigned a numerical rating, or a probable explanation for the blocks is that “SafeSurf Identification Standard” from 1-9. SafeSurf filtered out all unrated sites. Web publishers may assign themselves ratings in other categories as necessary – for example, SmartFilter a “nudity” rating of one if the site includes SmartFilter, manufactured by Secure Com- “subtle innuendo; [nudity] subtly implied puting, was originally intended for companies through the use of composition, lighting, seeking to limit their employees’ non-work- shaping, revealing clothing, etc.,” or a rating related Internet usage. By 1999, it was also of seven if it presents “erotic frontal nudity.” targeting schools.69 The filter’s control list has been modified, but on the whole, prior to In 1996, SafeSurf and 22 other companies 2001 SmartFilter divided objectionable sites founded the Platform for Internet Content into 27 categories, which could be enabled Selection (PICS). After Web publishers rate according to each customer’s needs. When their sites, PICS-compliant software can read SmartFilter 3.0 was unveiled in January 2001, the ratings and filter accordingly. SafeSurf ’s three of the categories (“alternative journals,” filter “is a server solution, which means that “non-essential,” and “worthless”) had been the software is not installed at the end user’s removed, and six others added, including computer, but at the ISP level to avoid tam- “mature” and “nudity.” pering.”68 In 2003, Secure Computing acquired Peacefire, “SafeSurf Examined” (2000) N2H2, and their databases were merged. In SafeSurf blocked multiple sites containing 2006, the SmartFilter Control List contained 68 “ Just the FAQs Please,” www.safesurf.com/ssfaq.htm (visited 3/10/06); see also “Microsoft Teams With Recreational 69 S ecure Computing, “Education and the Internet: A Bal- Software Advisory Council To Pioneer Parental Control Over anced Approach of Awareness, Policy, and Security” (1999), Internet Access,” www.w3.org/PICS/960228/Microsoft.html www.netapp.com/ftp/internet_and_education.pdf (visited (visited 2/26/06). 3/14/05). 32 Internet Filters: A Public Policy Report 73 content-based categories. The company access was filtered by SmartFilter. Censorware says that it uses “a combination of advanced deemed about 350 Web pages needlessly technology and highly skilled Web analysts” blocked under one or more of the five catego- to identify sites for blocking.70 ries chosen by UEN: “criminal skills,” “drugs,” “gambling,” “hate speech” and “sex.” Karen Schneider, A Practical Guide to Internet Filters (1997) Secure Computing claimed that “sites are not added to the Control List without first Testing SmartFilter with only its “sex” cat- being viewed and approved by our staff,” yet egory enabled, TIFAP found 12 sites blocked Censorware found that the home page of the – seven of them erroneously, in TIFAP’s Instructional Systems Program at Florida State estimation. These included three sites on University was blocked under the “gam- marijuana use, three gay-interest sites, and a bling” category, presumably because the word site containing safe sex information. “wager” appears in the URL. (Walter Wager, Peacefire, “SmartFilter Examined” (1997) a member of the program faculty, apparently maintained the site). SmartFilter also blocked Peacefire tested SmartFilter configured to “Marijuana: Facts for Teens,” a brochure block sites falling into categories likely to be published by the National Institute on Drug activated in a school setting: “criminal skills,” Abuse. Censorware’s findings also strongly “drugs,” “gambling,” “hate speech,” and “sex.” suggested that SmartFilter blocked the entire Among the sites blocked were: Community Wiretap server under the category of “crimi- United Against Violence, which works to pre- nal skills” – on account, it seems, of its URL vent anti-gay hate crime; Peaceable Texans for – even though Wiretap consists solely of Firearms Rights; the Marijuana Policy Project; electronic texts such as presidential inaugural the National Institute on Drug Abuse; Mother addresses, the Declaration of Independence, Jones magazine; the United States Information Shakespeare’s complete plays, The Jungle Book, Agency; the American Friends Service Com- Moby Dick, and the Book of Mormon. mittee; the Consortium on Peace Research, Education, and Development; the gay-themed Oasis Magazine; the Stop AIDS Project; and SmartFilter blocked Campaign for Our Children, a nonprofit or- “Marijuana: Facts for Teens,” ganization working to prevent teen pregnancy. SmartFilter also blocked sites containing edu- a brochure published by cational information on sexually transmitted the National Institute on diseases, safer sex, and teen pregnancy. Drug Abuse. Michael Sims et al., Censored Internet Access in Another server entirely blocked, for reasons Utah Public Schools and Libraries (Censorware unclear, was gopher.igc.apc.org, under the Project, Mar. 1999) “drugs” category; this server of the Institute The Censorware Project secured Internet log for Global Communications was home to files from Sept. 10-Oct. 10, 1998, of the Utah numerous nonprofit groups, such as the Rain- Education Network, or UEN, a state agency forest Action Network, Human Rights Watch, responsible for providing telecommunica- and Earth First. tions services to all the state’s public schools In other cases, possibly owing to keyword and many of its libraries. UEN’s Internet or URL-based filtering, pages were blocked whose aims were to raise awareness of such 70 “ SmartFilter Control List,” www.securecomputing.com/in- dex.cfm?skey=86 (visited 3/3/06). issues as hate seech and drugs. Among these Brennan Center for Justice 33 were Hate Watch, a site monitoring and was one ‘wrongly’ blocked access.” He also opposing online hate speech; a scholarly noted that Censorware’s investigation did paper titled “‘... as if I were the master of the not include sites on SmartFilter’s block list situation’: Proverbial Manipulation in Adolf that were overridden by the UEN – such as Hitler’s Mein Kampf,” from the archives of mormon.com, which accounted for 6,434 De Proverbio: An Electronic Journal of Interna- of the total 122,700 blocked page requests. tional Proverb Studies; the Iowa State Division “Counting these accesses,” he wrote, “would of Narcotics Enforcement; and a page on raise the error rate from 1 in 22 to 1 in 19.” the Web site of National Families in Action, In addition, the approximately 300 blocked a national drug education, prevention, and sites actually represented 5,601 individual policy center. wrongful blocks. Three months after Censored Internet Access Regarding Burt’s analyses of the sites was published, Secure Computing issued a deemed needlessly blocked, Censorware con- press release interpreting the report as a con- ceded that he was, in a few cases, correct (the firmation of SmartFilter’s effectiveness. Dur- sites in question being pornographic after all); ing the period in question, the release stated, yet Secure Computing actually removed them “there were over 54 million Web access at- from the SmartFilter database after the first tempts and of those, according to the report, report appeared – and added the Censorware less than 300 were denied access because the Project’s site, in all 27 blocking categories. site contacted had been miscategorized. This Seth Finkelstein, “SmartFilter’s Greatest Evils” represents stunning accuracy rate of 99.9994 (Nov. 16, 2000) percent.”71 Similarly, David Burt posted a report in which he claimed that only 279 sites Finkelstein found that SmartFilter blocked were actually included in Censorware’s study, a number of privacy and anonymous surfing after eliminating sites listed more than once, sites, many of which allow users to circumvent and that of these, only 64 actually constituted filtering software, in every category except inappropriate blocks. Burt, however, often “non-essential.” He named 19 such services, grouped as one erroneous block what actually including www.anonymizer.com; www. amounted to the blocking of multiple distinct freedom.net; www.private-server.com; and pages on a single server.72 www.silentsurf.com. SmartFilter also blocked, under every available classification but “non- Jamie McCarthy, “Lies, Damn Lies, and Sta- essential,” many sites providing translations tistics” (Censorware Project, June 23, 1999) of foreign language Web pages, for instance This was the Censorware Project’s response www.babelfish.org, www.onlinetrans.com, to Secure Computing’s and David Burt’s www.voycabulary.com, and www.worldlingo. claims that its study, Censored Internet Access, com. While such sites did not fall within demonstrated the accuracy of SmartFilter. SmartFilter’s published blocking criteria at the Author Jamie McCarthy stated that the actual time, they would very shortly thereafter, with overblocking rate was about 5% because, “for the introduction of SmartFilter 3.0. every 22 times SmartFilter ‘correctly’ blocked Seth Finkelstein, “SmartFilter – I’ve Got a someone from accessing a Web page, there Little List” (Dec. 7, 2000) 71 S ecure Computing press release, “Censorware Project Unequivocally Confirms Accuracy of SmartFilter in State of Finkelstein conducted a series of tests with Utah Education Network” (June 18, 1999). SmartFilter enabled to block only “extreme/ 72 David Burt, “Study of Utah School Filtering Finds ‘About obscene” material. Among the blocked sites 1 in a Million’ Web Sites Wrongly Blocked” (Filtering Facts, Apr. 4, 1999). were one for gay and lesbian Mormons; oth- 34 Internet Filters: A Public Policy Report
Enter the password to open this PDF file:
-
-
-
-
-
-
-
-
-
-
-
-