用户数据退出流程中的操纵性设计模式

Good LuckOpting Out Manipulative Design Patterns ABOUT EPIC The Electronic Privacy InformationCenter (EPIC) is a 501(c)(3) non-profit publicinterest research advocacy center in Washington, D.C. EPIC was establishedin 1994 to focus public attention on emerging privacy and civil liberties issues. AUTHORS Caroline Kraczon, EPICCounselJustin Sherman, EPIC Scholar In Residence ACKNOWLEDGEMENTS The authors would like to thank EPIC Counsel Kara Williams, Senior CounselSara Geoghegan, and Deputy Director and Policy Director Caitriona Fitzgeraldfor their comments on this report, as well as Deputy Director and Director ofEnforcement John Davisson for his support throughout this research effort. FUNDING STATEMENT This project was supported by funding from the Foundation for Public Service(FPS)—a 501(c)(3) fiscally sponsored project of Global Impact. HOW TO SUPPORT EPIC EPIC’s mission is to secure the fundamental right to privacy in the digital agefor all people through advocacy, research, and litigation. As a fully independent Please consider donating online atepic.org/donate.To learn about aboutadditional ways to support EPIC, such as donor-advised funds and qualified TABLE OF CONTENTS Introduction..............................................................................................................................4Literature Summary..............................................................................................................8Legal Summary......................................................................................................................14Methodology...........................................................................................................................17Analysis...................................................................................................................................23Overall Trends (Not Necessarily Manipulative)....................................................................23Manipulative Design Patterns...................................................................................................27Failing to Provide a Clear Mechanism to Opt Out of Sale and Sharing of PersonalInformation.................................................................................................................................27 Not Clearly Linking Opt-Out Form from Homepage and/or Privacy Policy.............29 Appendix A: State Privacy Law Language Regarding Opt-Outs..........................43 Introduction Consumers face an incredible power imbalancewhen trying to opt out of thecollection, use, transfer, or sale of their personal data. The average person usesmany online platforms and applications, meaning there are many companieswith which they directly interact that can store or transfer their data. On top ofthat, many other companies—including large language model (LLM) vendors thatscrape the internet and data brokers that collect people’s data without any actual, This problem is perhaps no clearer than in cases where individuals are at risk ofphysical violence, such as from online doxxing or stalking. Public servants across the country, at all levels of government, are facing an increase in violent threatsto themselves and their families.1In plenty of cases, there is a direct connectionto personal data exposure, such as when the individual who allegedly murderedMinnesota state legislator Melissa Hortman and her husband Mark in 2025—andshot and critically wounded another legislator and his wife—used “people search”data brokers to research his targets beforehand.2For decades, abusive individuals In this study, we evaluate whether major data-collecting companies’ opt-out pathsfor consumers utilize manipulative design practices. We do so by drawing on (FTC) enforcement actions, and our own experience to develop a list of specificdesign elements and behaviors that constitute manipulative design whilenavigating an opt-out process. Then, we apply those criteria to 38 companies:15 data brokers, four social media companies, nine other Big Tech companies, We find at least eight major manipulative design patterns across the 38 →Failing to provide a clear mechanism to opt out of sale and sharing of →Not clearly linking opt-out form from homepage and/or privacy policy →Requiring consumers to submit multiple separate forms →Deceptive statements about opt-outs and their success →Confusing or misleading language →Requiring consumers to log in or pay for a subscription before opting out →Design elements hiding important opt-out information (including designelements inducing false beliefs, hiding or delaying disclosure of material →Checkbox options preselected We also identified several other, overall trends (not necessarily manipulative).For example, many companies used the same vendor for their opt-out formor webpage; several companies, primarily the location data brokers, requiredconsumers to submit mobile ad identif