Real Safety AI Foundation / Research

Disability, Accessibility, and Civil Rights

The Human Signal: Reader-Facing AI Detection, the Disclosure Field, and the Tax on Assisted Authorship

Publisher: Real Safety AI Foundation

Working draft. Not peer reviewed.

Written
August 2026
Version
v0.5
Pages
27
Download PDFPlain language slides (PDF)Export BibTeXExport RIS

Abstract

In July 2026, Substack integrated a commercial AI detector into its reading surface, letting any reader scan any newly published post and receive an estimate of how much of the text a machine produced, displayed as a percentage beside the author's name. Announcing a funding round the following week, the detection company's chief executive stated the mission in a sentence: ``If we do not actively discriminate in favor of human content, then we're just gonna see more and more AI, and it's just gonna drown out any human signal that we have.'' This Article takes the sentence at its word and asks the question it leaves open: who counts as human when authorship runs through assistive technology. The instrument examined here is the third of its kind, after the invisible watermark and the crowdsourced authenticity flag, and it perfects what the first two began, because it ships with the remedy attached. Beside the detection badge sits a voluntary disclosure field where the author may explain how the work is made. The Article's central claim is that the pairing converts the field into a tax assessed by disability. A nondisabled writer clears an adverse badge with a style preference; a writer whose work runs through dictation software and a language model clears it only by publishing a diagnosis. The exculpatory answer is priced differently for the two populations, silence becomes an answer once the field exists, and rising detector accuracy raises the tax rather than curing it, because the harm was never the error rate. The harm is a taxonomy with no category for assisted authorship, enforced by an increasingly reliable instrument. The Article then shows that every accommodation-shaped repair fails a one-sentence test, since a badge, an exemption tag, or a visible opt-out that permits any inference of disability is itself the disclosure, and it specifies the universal-design remedies that pass. It closes with the doctrine: the screen-out and methods-of-administration provisions of ADA Title III, and the New York City Human Rights Law, which reaches the detection company where it lives.

Keywords

  • content provenance
  • assistive technology
  • compelled disclosure
  • disability discrimination
  • disparate impact
  • universal design
  • public accommodation
  • screen-out
  • disclosure penalty
  • New York City Human Rights Law
  • platform governance
  • forced intimacy

Plain language slides

First slide of the plain language summary of The Human Signal: Reader-Facing AI Detection, the Disclosure Field, and the Tax on Assisted AuthorshipOpen the 27-slide summary (PDF)

Suggested citation

Gilly, Travis. "The Human Signal: Reader-Facing AI Detection, the Disclosure Field, and the Tax on Assisted Authorship." Real Safety AI Foundation Working Draft, August 2026. https://realsafetyai.org/research/pykm9c/

Other versions

This paper is also posted on SSRN.

SSRN version

References (37)

  1. Americans with Disabilities Act of 1990, 42 U.S.C. §§ 12101–12213.
  2. Bellan, R. (2026). As AI content floods the internet, Pangram raises $9M to detect it. TechCrunch. https://techcrunch.com/2026/07/29/as-ai-content-floods-the-internet-pangram-raises-9m-to-detect-it/.
  3. Bonifacic, I. (2026). Substack is adding an AI detection feature. Engadget. https://www.engadget.com/2220064/substack-is-adding-an-ai-detection-feature/.
  4. Carparts Distribution Center, Inc. v. Automotive Wholesaler’s Association of New England, Inc., 37 F.3d 12 (1st Cir. 1994).
  5. Coley, M. (2023). Guidance on AI detection and why we’re disabling Turnitin’s AI detector. Vanderbilt University Brightspace. https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/.
  6. Doe v. Mutual of Omaha Insurance Co., 179 F.3d 557 (7th Cir. 1999).
  7. Fair Housing Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008) (en banc).
  8. Gedzyk-Nieman, S. A., & Derouin, A. (2025). Ensuring access for all nursing students: Classroom accommodations and universal design for learning. Creative Nursing, 31(4), 392–398. https://doi.org/10.1177/10784535251392561.
  9. Kendro, K., Maloney, J., & Jarvis, S. (2026). Do large language models produce texts with “human-like” lexical diversity? Evidence from four ChatGPT models. International Journal of Applied Linguistics. https://doi.org/10.1111/ijal.70115.
  10. Lawrence, K. W., Habibi, A. A., Ward, S. A., Lajam, C. M., Schwarzkopf, R., & Rozell, J. C. (2024). Human versus artificial intelligence-generated arthroplasty literature: A single-blinded analysis of perceived communication, quality, and authorship source. The International Journal of Medical Robotics and Computer Assisted Surgery, 20(1), e2621. https://doi.org/10.1002/rcs.2621.
  11. Elkhatat, A. M., Elsaid, K., & Almeer, S. (2023). Evaluating the efficacy of AI content detection tools in differentiating between human and AI-generated text. International Journal for Educational Integrity, 19, art. 17. https://doi.org/10.1007/s40979-023-00140-5.
  12. European Commission. (2026). Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act. https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems.
  13. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. https://doi.org/10.1016/j.patter.2023.100779.
  14. Liçenji, L., & Hoxha, J. (2026). When news is “written by artificial intelligence”: A systematic review of provenance and disclosure cues in journalism and their effects on credibility and trust. Frontiers in Artificial Intelligence, 9, art. 1815243. https://doi.org/10.3389/frai.2026.1815243.
  15. Moloney, M., Brown, R. L., & Ciciurkaite, G. (2018). “Going the extra mile”: Disclosure, accommodation, and stigma management among working women with disabilities. Deviant Behavior, 40(8), 942–956. https://doi.org/10.1080/01639625.2018.1445445.
  16. Moody v. NetChoice, LLC, 603 U.S. 707 (2024).
  17. Moreland, C. J., Pereira-Lima, K., Lee, K. T., et al. (2026). Disability accommodation access and requests in US internal medicine residents with disabilities. JAMA Network Open, 9(3), e263392. https://doi.org/10.1001/jamanetworkopen.2026.3392.
  18. Nakano, H., Takezawa, J., & Matulic, F. (2025). Understanding reader perception shifts upon disclosure of AI authorship. arXiv. https://doi.org/10.48550/arxiv.2510.24011.
  19. Mayor, E., Bietti, L. M., & Bangerter, A. (2025). Can large language models simulate spoken human conversations? Cognitive Science, 49(9), e70106. https://doi.org/10.1111/cogs.70106.
  20. Mingus, M. (2017). Forced intimacy: An ableist norm. Leaving Evidence. https://leavingevidence.wordpress.com/2017/08/06/forced-intimacy-an-ableist-norm/.
  21. New York City Human Rights Law, N.Y.C. Admin. Code §§ 8-101 to 8-131.
  22. Pangram Labs. (2025). AI transparency company Pangram integrated into two leading news source and resource platforms [Press release]. Business Wire. https://www.businesswire.com/news/home/20250915315368/en.
  23. Pereira-Lima, K., Meeks, L. M., Ross, K. E. T., et al. (2023). Barriers to disclosure of disability and request for accommodations among first-year resident physicians in the US. JAMA Network Open, 6(5), e239981. https://doi.org/10.1001/jamanetworkopen.2023.9981.
  24. PGA Tour, Inc. v. Martin, 532 U.S. 661 (2001).
  25. Raj, M., Berg, J. M., & Seamans, R. (2026). The artificial intelligence disclosure penalty: Humans persistently devalue AI-generated creative writing. Journal of Experimental Psychology: General, 155(4), 896–915. https://doi.org/10.1037/xge0001889.
  26. Robles v. Domino’s Pizza, LLC, 913 F.3d 898 (9th Cir. 2019).
  27. Schilke, O., & Reimann, M. (2025). The transparency dilemma: How AI disclosure erodes trust. Organizational Behavior and Human Decision Processes, 188, 104405. https://doi.org/10.1016/j.obhdp.2025.104405.
  28. Perez, S. (2026). Substack’s new tool tells you who’s been writing their newsletters with AI. TechCrunch. https://techcrunch.com/2026/07/22/substacks-new-tool-tells-you-whos-been-writing-their-newsletters-with-ai/.
  29. Regulation (EU) 2024/1689 (Artificial Intelligence Act), art. 50(2), 2024 O.J. (L 1689).
  30. Substack. (2026). Announcement of Pangram AI detection integration. X. https://x.com/Substack/status/2079598704424779787.
  31. Communications Decency Act § 230, 47 U.S.C. § 230.
  32. Ullah, U., Laudanna, S., Di Sorbo, A., Visaggio, C. A., & Murray, R. (2026). Breaking the imitation game: Can LLMs fool humans and machines alike? International Journal of Intelligent Systems, 2026(1), 8117975. https://doi.org/10.1155/int/8117975.
  33. Wang, N., & Liang, C. (2026). How to disclose? Strategic AI disclosure in crowdfunding. arXiv. https://doi.org/10.48550/arxiv.2602.15698.
  34. U.S. Department of Justice, Nondiscrimination on the Basis of Disability by Public Accommodations, 28 C.F.R. pt. 36.
  35. Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Šigut, P., & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19, art. 26. https://doi.org/10.1007/s40979-023-00146-z.
  36. Weyer v. Twentieth Century Fox Film Corp., 198 F.3d 1104 (9th Cir. 2000).
  37. Zhu, X., Jiang, F., & Volpone, S. D. (2025). The role of inclusive leadership in reducing disability accommodation request withholding. Journal of Management, 52(4), 1570–1601. https://doi.org/10.1177/01492063251314453.

All research