Real Safety AI Foundation / Research

Disability, Accessibility, and Civil Rights

Reliable Enough to Punish: K-12 Surveillance AI and the Section 504 Evaluation the District Insists It Is Not Conducting

Publisher: Real Safety AI Foundation

Working draft. Not peer reviewed.

Written
August 2026
Version
v0.6
Pages
26
Download PDFPlain language slides (PDF)Export BibTeXExport RIS

Abstract

A majority of American public school students now pass their school days, and much of their time at home, inside software that reads what they type, watches what they search, and scores what it sees. The vendors call the products safety monitoring. One of them now assigns each child a wellness level. The districts that buy these systems treat their outputs as reliable when the output supports a suspension and treat the same outputs as too informal to count when the output would obligate the district to do anything. This Article argues that the second half of that position is no longer available to them. The argument proceeds from a fork the districts built for themselves. Section 504's child find and evaluation regulations oblige a district to evaluate any student it has reason to believe needs special education or related services because of disability, with notice and procedural safeguards attached. A district that disciplines a student on the strength of an algorithmic flag has stated, in its own records, that it believes the flag; a district that believes a flag describing self harm, emotional disturbance, or psychological crisis has acquired exactly the belief that triggers the evaluation duty it did not follow. The district may characterize its system as reliable or as unreliable. It may not select the characterization by consequence. Beneath the fork the Article locates three further failures. Continuous behavioral scoring directed at mental and psychological states is an evaluation under the Protection of Pupil Rights Amendment and bears every operative mark of a medical examination under the criteria courts already apply, no clinician required, no diagnosis required, the practical effect controlling. A scoring system that flags students for perceived psychological abnormality manufactures regarded as claimants at the scale of enrollment. And a compulsory education delivered only through monitored platforms denies meaningful access to the students the system systematically misreads, with students on the psychotic spectrum injured first and worst, because announced continuous surveillance removes the one proposition their treatment requires them to be able to test. The constitutional challenges now in the courts are failing on qualified immunity for want of precedent. The disability statutes have been clearly established for forty years, they run against the entity, and nobody has pled them.

Keywords

  • Section 504
  • Title II
  • student surveillance
  • Gaggle
  • wellness scoring
  • child find
  • medical examination
  • regarded as
  • Alexander v. Choate
  • Protection of Pupil Rights Amendment
  • qualified immunity

Plain language slides

First slide of the plain language summary of Reliable Enough to Punish: K-12 Surveillance AI and the Section 504 Evaluation the District Insists It Is Not ConductingOpen the 31-slide summary (PDF)

Suggested citation

Gilly, Travis. "Reliable Enough to Punish: K-12 Surveillance AI and the Section 504 Evaluation the District Insists It Is Not Conducting." Real Safety AI Foundation Working Draft, August 2026. https://realsafetyai.org/research/reliable-enough-to-punish/

Other versions

This paper is also posted on SSRN.

SSRN version

References (66)

This paper cites its sources in footnotes. Each authority is listed once, where it is first cited, with its footnote number.

  1. Footnote 1.Tell v. Lawrence Bd. of Educ., No. 2:25-cv-02428 (D. Kan. Apr. 8, 2026).
  2. Footnote 4.29 U.S.C. § 794(a).
  3. Footnote 5.42 U.S.C. § 12132.
  4. Footnote 6.Alexander v. Choate, 469 U.S. 287, 301 (1985).
  5. Footnote 8.Barbara Fedders, The Constant and Expanding Classroom: Surveillance in K-12 Public Schools, 97 N.C. L. Rev. 1673 (2019)
  6. Footnote 8.Danielle Keats Citron, The Surveilled Student, 76 Stan. L. Rev. 1439 (2024).
  7. Footnote 9.Wellness Levels in Securly Aware, Securly (Mar. 25, 2024), https://blog.securly.com/wellness-levels-securly-aware-holistic-support/
  8. Footnote 9.Why Securly Aware Is a More Effective & Reliable Way To Monitor Student Wellness, Securly (Apr. 10, 2025), https://blog.securly.com/more-effective-student-wellness-monitoring/.
  9. Footnote 9.Nicholas E. Stewart, We’ve Replaced School Security With the Creepiest Possible System, Slate (Aug. 21, 2026), https://slate.com/news-and-politics/2026/08/back-to-school-security-creepy-ai-surveillance-system.html.
  10. Footnote 10.34 C.F.R. § 104.35(a).
  11. Footnote 11.34 C.F.R. § 104.36.
  12. Footnote 12.Ctr. for Democracy & Tech., Hidden Harms: The Misleading Promise of Monitoring Students Online (2022).
  13. Footnote 13.Elizabeth Warren & Ed Markey, Constant Surveillance: Implications of Around-the-Clock Online Student Activity Monitoring (2022), https://perma.cc/5GH5-HATU.
  14. Footnote 14.Complaint, Merrill v. Marana Unified Sch. Dist. (D. Ariz.), https://storage.courtlistener.com/recap/gov.uscourts.azd.1450446/gov.uscourts.azd.1450446.1.0.pdf
  15. Footnote 14.EFF to Arizona Federal Court: Protect Public School Students from Surveillance and Punishment for Off-Campus Speech, Elec. Frontier Found. (Nov. 26, 2025), https://www.eff.org/deeplinks/2025/11/eff-arizona-federal-court-protect-public-school-students-surveillance-and (amicus brief arguing that use of a school-issued laptop or account does not categorically place a student on campus, and noting that the district runs Gaggle, GoGuardian, and Securly together on its Chromebooks and Google Workspace).
  16. Footnote 18.Lynsay Ayer et al., RAND Corp., Artificial Intelligence-Based Student Activity Monitoring for Suicide Risk: Considerations for K-12 Schools, Caregivers, Government, and Technology Developers (2023), https://doi.org/10.7249/rra2910-1.
  17. Footnote 19.Dhanaraj Thakur, Hugh Grant-Chapman & Elizabeth Laird, Ctr. for Democracy & Tech., Beyond the Screen: Parents’ Experiences with Student Activity Monitoring in K-12 Schools (2023), https://doi.org/10.31219/osf.io/bwjc4
  18. Footnote 19.DeVan L. Hankerson, Cody Venzke & Elizabeth Laird, Ctr. for Democracy & Tech., Online and Observed: Student Privacy Implications of School-Issued Devices and Student Activity Monitoring Software (2022), https://doi.org/10.31219/osf.io/73a2b (eighty percent figure; documenting that community concerns centered on “appropriate use of student activity monitoring data for disciplinary purposes”).
  19. Footnote 23.Student Wellness Monitoring Solution: Securly Aware, Securly, https://www.securly.com/aware (“At-Risk AI conducts a nuanced analysis of flagged events, including a student’s activity history, to determine a student’s Wellness Level”; describing the teacher AI assistant and the pop-up of mental health resources “if their wellness levels decline”).
  20. Footnote 26.29 U.S.C. § 794(b).
  21. Footnote 28.28 C.F.R. § 35.130(b)(8).
  22. Footnote 29.28 C.F.R. § 35.130(b)(7)(i).
  23. Footnote 30.28 C.F.R. § 35.130(f).
  24. Footnote 31.28 C.F.R. § 35.160(a)(1).
  25. Footnote 32.105 Ill. Comp. Stat. 5/26-1 (whoever has custody or control of a child between six and seventeen “shall cause such child to attend some public school in the district wherein the child resides the entire time it is in session,” subject to enumerated exemptions).
  26. Footnote 38.20 U.S.C. § 1232h(b)(2).
  27. Footnote 39.Gonzaga Univ. v. Doe, 536 U.S. 273 (2002) (no private right of action to enforce FERPA’s nondisclosure provisions under § 1983).
  28. Footnote 40.C.N. v. Ridgewood Bd. of Educ., 430 F.3d 159 (3d Cir. 2005).
  29. Footnote 41.Karraker v. Rent-A-Center, Inc., 411 F.3d 831, 835 (7th Cir. 2005).
  30. Footnote 48.42 U.S.C. § 12112(d).
  31. Footnote 49.Mobley v. Workday, Inc., No. 3:23-cv-00770 (N.D. Cal. July 1, 2026).
  32. Footnote 52.Zell v. Ricci, 321 F. Supp. 3d 285 (D.R.I. 2018) (concluding that neither a five day suspension nor an in school detention implicated the statutes’ protections) (citing Honig v. Doe, 484 U.S. 305, 325 (1988)).
  33. Footnote 53.42 U.S.C. § 12102(3)(A).
  34. Footnote 54.Lawton v. Success Acad. Charter Schs., Inc., 323 F. Supp. 3d 353 (E.D.N.Y. 2018) (quoting the perceived impairment language)
  35. Footnote 54.Voss v. Baca, No. 3:14-cv-00066 (D. Nev. July 9, 2015)
  36. Footnote 54.Widomski v. Orange Cnty. Cmty. Coll., No. 7:09-cv-07517 (S.D.N.Y. Mar. 21, 2013).
  37. Footnote 55.42 U.S.C. § 12102(3)(B) (“Paragraph (1)(C) shall not apply to impairments that are transitory and minor. A transitory impairment is an impairment with an actual or expected duration of 6 months or less.”).
  38. Footnote 56.Nipper v. Harrison Sch. Dist. Two, No. 1:17-cv-02391 (D. Colo. Apr. 15, 2020).
  39. Footnote 57.Chadam v. Palo Alto Unified Sch. Dist., No. 4:13-cv-04129 (N.D. Cal. Nov. 4, 2014) (quoting § 12102’s regarded as definition in a suit over a district’s response to genetic information in student records).
  40. Footnote 58.28 C.F.R. § 35.130(b)(7)(ii).
  41. Footnote 61.Payan v. Los Angeles Cmty. Coll. Dist., 11 F.4th 729, 736–39 (9th Cir. 2021) (recognizing the theory), with Doe v. BlueCross BlueShield of Tenn., Inc., 926 F.3d 235, 241 (6th Cir. 2019).
  42. Footnote 62.Daniel Freeman et al., Virtual Reality in the Treatment of Persecutory Delusions: Randomised Controlled Experimental Study Testing How To Reduce Delusional Conviction, 209 Brit. J. Psychiatry 62, 62 (2016), https://doi.org/10.1192/bjp.bp.115.176438.
  43. Footnote 62.Daniel Freeman et al., An Explanatory Randomised Controlled Trial Testing the Effects of Targeting Worry in Patients with Persistent Persecutory Delusions: The Worry Intervention Trial, 2 Efficacy & Mechanism Evaluation 1 (2015), https://doi.org/10.3310/eme02010.
  44. Footnote 63.DOE v. Township High Sch. Dist. No. 214, No. 1:19-cv-03052 (N.D. Ill. Aug. 21, 2024) (student journaled “patterns and lists in his notebook as a way of coping with his anxiety”; the school’s response to the notebook generated the Title II litigation).
  45. Footnote 64.R.W. v. Bd. of Regents of the Univ. Sys. of Ga., 114 F. Supp. 3d 1260 (N.D. Ga. 2015) (campus threat assessment premised on links between mental health and violence)
  46. Footnote 64.Ogletree v. Cleveland State Univ., 647 F. Supp. 3d 602 (N.D. Ohio 2022) (federal challenge to remote proctoring surveillance, including Honorlock, at a public university).
  47. Footnote 66.34 C.F.R. § 99.3.
  48. Footnote 67.34 C.F.R. § 99.31(a)(1)(i)(B).
  49. Footnote 68.34 C.F.R. § 99.33(a).
  50. Footnote 69.34 C.F.R. § 99.8(b)(1) (records of a law enforcement unit are those “created by that law enforcement unit,” created “for a law enforcement purpose,” and “[m]aintained by the law enforcement unit”)
  51. Footnote 70.34 C.F.R. § 99.8(d).
  52. Footnote 75.Cummings v. Premier Rehab Keller, P.L.L.C., 596 U.S. 212 (2022).
  53. Footnote 76.Payan v. Los Angeles Cmty. Coll. Dist., 169 F.4th 971, 977–79 (9th Cir. 2024).
  54. Footnote 77.20 U.S.C. § 1415(l)
  55. Footnote 77.Fry v. Napoleon Cmty. Schs., 580 U.S. 154 (2017) (exhaustion turns on whether the gravamen of the complaint is the denial of a FAPE).
  56. Footnote 78.Luna Perez v. Sturgis Pub. Schs., 598 U.S. 142 (2023) (suit seeking compensatory damages under the ADA may proceed without exhausting IDEA procedures because the relief is unavailable under the IDEA).
  57. Footnote 79.Duvall v. County of Kitsap, 260 F.3d 1124, 1139 (9th Cir. 2001) (damages under Title II require intentional discrimination, shown by deliberate indifference: knowledge that a harm to a federally protected right is substantially likely, and a failure to act).
  58. Footnote 81.Shanahan v. IXL Learning, Inc., No. 3:24-cv-02724 (N.D. Cal. Nov. 1, 2024) (even if districts had obtained the required parental authorization for data collection, “that still does not mean the parents impliedly authorized” the vendor’s further practices).
  59. Footnote 82.Myers v. Beaverton Sch. Dist. 48J, No. 3:25-cv-00677 (D. Or. Apr. 17, 2026) (published opinion reciting allegations that data reviews “can result in staff-initiated meetings with students, without prior notice to parents,” and that while parents can view traditional attendance data through the parent portal, “parents cannot view the data collected through the DHP system”).
  60. Footnote 83.Fitzpatrick v. Town of Falmouth, 2005 ME 97, 879 A.2d 21
  61. Footnote 83.Sch. Bd. of Nassau Cty. v. Arline, 480 U.S. 273 (1987) (the significant risk inquiry is individualized and grounded in objective evidence rather than in fears or stereotypes about disability).
  62. Footnote 85.APDCAT (Catalonia), Decision No. PS 49/2019 (fingerprint and facial recognition attendance control in a public secondary school), summarized at https://gdprhub.eu/index.php?title=APDCAT_(Catalonia)_-_PS_49/2019.
  63. Footnote 86.Datainspektionen (Swed.), Decision No. DI-2019-7024 (Nov. 24, 2020) (Stockholm “Skolplattformen” processing personal data of 500,000 students), summarized at https://gdprhub.eu/index.php?title=Datainspektionen_-_DI-2019-7024.
  64. Footnote 87.AEPD (Spain), Decision No. EXP202200367 (facial recognition identity confirmation for online examinations), summarized at https://gdprhub.eu/index.php?title=AEPD_(Spain)_-_EXP202200367.
  65. Footnote 88.Persónuvernd (Ice.), Decision No. 2020112830 (Mar. 8, 2022), summarized at https://gdprhub.eu/index.php?title=Persónuvernd_(Iceland)_-_2020112830.
  66. Footnote 89.Anna Carlile, School Surveillance, Control, and Resistance in the United Kingdom, in The Palgrave International Handbook of School Discipline, Surveillance, and Social Control 17 (2018), https://doi.org/10.1007/978-3-319-71559-9_2.

All research