Real Safety AI Foundation / Research

Disability, Accessibility, and Civil Rights

The One-Way Valve: Crowdsourced Authenticity Flagging, Assistive Technology, and Disability Discrimination on the Professional Platform

Publisher: Real Safety AI Foundation

Working draft. Not peer reviewed.

Written
August 2026
Version
v0.9
Pages
30
Download PDFPlain language slides (PDF)Export BibTeXExport RIS

Abstract

On July 30, 2026, LinkedIn shipped a report option labeled ``Seems like AI slop,'' inviting members to flag posts that read as machine-made, reducing the distribution of flagged posts, and feeding the flags into the platform's classifiers as training signal. This Article identifies the instrument's defining defect: it is a one-way valve. It collects accusations and nothing else. No countervailing judgment enters, ground truth never enters, and the resulting classifier learns complaint prevalence rather than authorship. The empirical record supplies each layer of the resulting injury. Humans cannot reliably distinguish machine text from human text, and their confidence bears no relation to their accuracy; crowd raters carry measured demographic biases that trained models absorb and amplify; evaluators penalize disclosed AI assistance even where the work is accurate and the author is human; and the writers most likely to be flagged write in the disciplined registers associated with disability, second-language acquisition, and assistive technology. The Article then maps the doctrine: disparate impact and meaningful access under the Americans with Disabilities Act and Section 504, the surcharge prohibition as applied to paid reach restoration, a procedural modification remedy that requires no alteration of any third party's content, and the resulting path through Section 230 in every circuit configuration, including the Fifth Circuit's July 2026 holding that the statutory shield and the First Amendment shield stack.

Keywords

  • AI text detection
  • artificial intelligence disclosure penalty
  • annotator bias
  • algorithmic amplification
  • content moderation
  • disparate impact
  • reasonable modification
  • surcharge doctrine
  • platform accountability
  • Section 230
  • speech-to-text dictation
  • professional networking platforms

Plain language slides

First slide of the plain language summary of The One-Way Valve: Crowdsourced Authenticity Flagging, Assistive Technology, and Disability Discrimination on the Professional PlatformOpen the 21-slide summary (PDF)

Suggested citation

Gilly, Travis. "The One-Way Valve: Crowdsourced Authenticity Flagging, Assistive Technology, and Disability Discrimination on the Professional Platform." Real Safety AI Foundation Working Draft, August 2026. https://realsafetyai.org/research/erv8r8/

Other versions

This paper is also posted on SSRN.

SSRN version

References (71)

  1. A.B. v. Salesforce, Inc., 123 F.4th 788 (5th Cir. 2024).
  2. Al Ali, A., Helcl, J., & Libovický, J. (2026). Different time, different language: Revisiting the bias against non-native speakers in GPT detectors (arXiv:2602.05769).
  3. Al Kuwatly, H., Wich, M., & Groh, G. (2020). Identifying and measuring annotator bias based on annotators’ demographic characteristics. Proceedings of the Fourth Workshop on Online Abuse and Harms, 184.
  4. Alexander v. Choate, 469 U.S. 287 (1985).
  5. Alexander v. Sandoval, 532 U.S. 275 (2001).
  6. Altay, S., & Gilardi, F. (2024). People are skeptical of headlines labeled as AI-generated, even if true or human-made, because they assume full AI automation. PNAS Nexus, 3(10), pgae403.
  7. Americans with Disabilities Act of 1990, 42 U.S.C. §§ 12101–12213.
  8. Anderson v. TikTok, Inc., 116 F.4th 180 (3d Cir. 2024).
  9. Anthropic Help Center. (2026). Text watermarking on Claude models. https://support.claude.com/en/articles/16266773.
  10. Barnes v. Yahoo!, Inc., 570 F.3d 1096 (9th Cir. 2009).
  11. Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5, 323.
  12. Bordalejo, B., et al. (2025). “Scarlet Cloak and the Forest Adventure”: A preliminary study of the impact of AI on commonly used writing tools. International Journal of Educational Technology in Higher Education, 22(1), art. 6.
  13. Carparts Distribution Center, Inc. v. Automotive Wholesaler’s Association of New England, Inc., 37 F.3d 12 (1st Cir. 1994).
  14. Chambers, S., & Kelley, M. C. (2025). The misclassification of autistic writing as AI-generated. In A. I. Cristea et al. (Eds.), Artificial intelligence in education: 26th international conference, AIED 2025, proceedings, part III. Springer.
  15. Clark, E., et al. (2021). All that’s “human” is not gold: Evaluating human evaluation of generated text. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics, 7282.
  16. Communications Decency Act § 230, 47 U.S.C. § 230.
  17. Computer & Communications Industry Association v. Paxton, No. 24-50721 (5th Cir. July 24, 2026).
  18. Crawford, K., & Gillespie, T. (2016). What is a flag for? Social media reporting tools and the vocabulary of complaint. New Media & Society, 18, 410.
  19. Cummings v. Premier Rehab Keller, P.L.L.C., 596 U.S. 212 (2022).
  20. Doe 1 v. Meta Platforms, Inc., No. 24-1672 (9th Cir. Apr. 28, 2026).
  21. Doe v. BlueCross BlueShield of Tennessee, Inc., 926 F.3d 235 (6th Cir. 2019).
  22. Doe v. Grindr Inc., 128 F.4th 1148 (9th Cir. 2025).
  23. Doe v. Mutual of Omaha Insurance Co., 179 F.3d 557 (7th Cir. 1999).
  24. Emi, B., et al. (2024). Technical report on the Pangram AI-generated text classifier (arXiv:2402.14873). Pangram Labs.
  25. Estate of Bride v. Yolo Technologies, Inc., 112 F.4th 1168 (9th Cir. 2024).
  26. Federal Grant and Cooperative Agreement Act, 31 U.S.C. §§ 6301–6309.
  27. Federal Trade Commission Act § 5, 15 U.S.C. § 45.
  28. Ford v. Schering-Plough Corp., 145 F.3d 601 (3d Cir. 1998).
  29. GEO Group, Inc. v. Menocal, 607 U.S. 438 (2026).
  30. Gonzalez v. Google LLC, 598 U.S. 617 (2023) (per curiam).
  31. Haupt, M., Freidank, J., & Haas, A. (2024). Consumer responses to human-AI collaboration at organizational frontlines: Strategies to escape algorithm aversion in content creation. Review of Managerial Science, 19(2), 377–413.
  32. HomeAway.com, Inc. v. City of Santa Monica, 918 F.3d 676 (9th Cir. 2019).
  33. Illinois Consumer Fraud and Deceptive Business Practices Act, 815 ILCS 505/1 et seq.
  34. In re Apple Inc. App Store Simulated Casino-Style Games Litigation, 625 F. Supp. 3d 971 (N.D. Cal. 2022).
  35. Jacobson v. Delta Airlines, Inc., 742 F.2d 1202 (9th Cir. 1984).
  36. Jiang, Y., et al. (2024). Detecting ChatGPT-generated essays in a large-scale writing assessment: Is there a bias against non-native English speakers? Computers & Education, 224.
  37. Liang, W., et al. (2023). GPT detectors are biased against non-native English writers. Patterns, 4, 100779.
  38. Lloyd, T., et al. (2025). AI rules? Characterizing Reddit community policies towards AI-generated content. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems.
  39. Moody v. NetChoice, LLC, 603 U.S. 707 (2024).
  40. Mozafari, M., Farahbakhsh, R., & Crespi, N. (2020). Hate speech detection and racial bias mitigation in social media based on BERT model. PLoS ONE, 15, e0237861.
  41. Nakano, H., et al. (2025). Understanding reader perception shifts upon disclosure of AI authorship. Proceedings of the 31st International Conference on Intelligent User Interfaces.
  42. National Association of the Deaf v. Harvard University, 377 F. Supp. 3d 49 (D. Mass. 2019).
  43. NCAA v. Smith, 525 U.S. 459 (1999).
  44. Nondiscrimination on the Basis of Disability by Public Accommodations, 28 C.F.R. § 36.301.
  45. Nondiscrimination on the Basis of Disability in State and Local Government Services, 28 C.F.R. § 35.130.
  46. Parker v. Metropolitan Life Insurance Co., 121 F.3d 1006 (6th Cir. 1997) (en banc).
  47. Payan v. Los Angeles Community College District, 11 F.4th 729 (9th Cir. 2021).
  48. Payan v. Los Angeles Community College District, 169 F.4th 971 (9th Cir. 2026).
  49. Personal Injury Plaintiffs v. TikTok, LLC, No. 24-7312 (9th Cir. Aug. 10, 2026).
  50. PGA Tour, Inc. v. Martin, 532 U.S. 661 (2001).
  51. Prajod, P., Cools, H., & Röggla, T. (2026). Full disclosure, less trust? How the level of detail about AI use in news writing affects readers’ trust (arXiv:2601.09620).
  52. Raj, M., et al. (2026). The artificial intelligence disclosure penalty: Humans persistently devalue AI-generated creative writing. Journal of Experimental Psychology: General, 155(4), 896–915.
  53. Regulation (EU) 2022/2065 (Digital Services Act), 2022 O.J. (L 277) 1.
  54. Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5, 296.
  55. Rehabilitation Act of 1973 § 504, 29 U.S.C. § 794.
  56. Reif, J. A., Larrick, R. P., & Soll, J. B. (2025). Evidence of a social evaluation penalty for using AI. Proceedings of the National Academy of Sciences, 122(19), e2426766122.
  57. Robles v. Domino’s Pizza, LLC, 913 F.3d 898 (9th Cir. 2019).
  58. Sap, M., et al. (2019). The risk of racial bias in hate speech detection. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 1668.
  59. Schilke, O., & Reimann, M. (2025). The transparency dilemma: How AI disclosure erodes trust. Organizational Behavior and Human Decision Processes, article 104405.
  60. Sikhs for Justice, Inc. v. Facebook, Inc., 144 F. Supp. 3d 1088 (N.D. Cal. 2015).
  61. Toff, B., & Simon, F. M. (2024). “Or they could just not use it?”: The dilemma of AI disclosure for audience trust in news. The International Journal of Press/Politics, 30(4), 881–903.
  62. Ullah, U., Laudanna, S., Di Sorbo, A., & Visaggio, C. A. (2026). Breaking the imitation game: Can LLMs fool humans and machines alike? International Journal of Intelligent Systems, 2026, art. 8117975.
  63. U.S. Department of Transportation v. Paralyzed Veterans of America, 477 U.S. 597 (1986).
  64. Waltzer, T., Cox, R. L., & Heyman, G. D. (2023). Testing the ability of teachers and students to differentiate between essays generated by ChatGPT and high school students. Human Behavior and Emerging Technologies, 2023, art. 1923981.
  65. Waseem, Z. (2016). Are you a racist or am I seeing things? Annotator influence on hate speech detection on Twitter. Proceedings of the First Workshop on NLP and Computational Social Science, 138.
  66. Weber-Wulff, D., et al. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19.
  67. Wetzel v. Glen St. Andrew Living Community, LLC, 901 F.3d 856 (7th Cir. 2018).
  68. Weyer v. Twentieth Century Fox Film Corp., 198 F.3d 1104 (9th Cir. 2000).
  69. Wich, M., Al Kuwatly, H., & Groh, G. (2020). Investigating annotator bias with a graph-based approach. Proceedings of the Fourth Workshop on Online Abuse and Harms, 191.
  70. Xia, M., Field, A., & Tsvetkov, Y. (2020). Demoting racial bias in hate speech detection. Proceedings of the Eighth International Workshop on Natural Language Processing for Social Media, 7.
  71. Zhang, Y., & Gosline, R. (2023). Human favoritism, not AI aversion: People’s perceptions (and bias) toward generative AI, human experts, and human–GAI collaboration in persuasive content generation. Judgment and Decision Making, 18, e41.

All research