In a 48-platform policy audit, data-collection disclosure was strongest while AI and accountability disclosures were weak¶

Claim · All claims
Evidence · 1 study (3 entries) · qualitative r3 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Thirty-three percent of platforms (16/48) made no meaningful AI disclosure despite visible AI features. → Meghna Manoj Nair 2026 q2 i? Seventy-three percent of platforms received only a middle accountability score with generic breach language. → Meghna Manoj Nair 2026 (2) q2 i? Data-collection disclosure scored highest of the five dimensions (mean 1.81 of 2). → Meghna Manoj Nair 2026 (3)

Evidence¶

Meghna Manoj Nair 2026¶

Meghna Manoj Nair, Rachel Greenstadt. (2026). "We'll Fix It Later": Education, AI, and the Deferral of Student Privacy in EdTech. New York University. https://arxiv.org/abs/2609.28137 doi:10.48550/arxiv.2609.28137

q2 · i? · qualitative · r3

Audit finding from the Results section: two coders scored 48 platforms on a five-dimension rubric (mean Cohen's kappa 0.782), with AI disclosure (C4) among the lowest-scoring dimensions (mean 0.90 of 2).

"Thirty-three percent of platforms (16/48) made no meaningful AI disclosure despite visible AI features."

Meghna Manoj Nair 2026 (2)¶

Meghna Manoj Nair, Rachel Greenstadt. (2026). "We'll Fix It Later": Education, AI, and the Deferral of Student Privacy in EdTech. New York University. https://arxiv.org/abs/2609.28137

q2 · i? · qualitative · r3

Audit finding from the Results section. Accountability (C5) mean was 1.07 of 2, with 8% scoring 0, 73% scoring 1, and 19% scoring 2, indicating mostly generic rather than operational accountability disclosure.

"Accountability was similarly shallow: 73% received the middle score, typically a generic email address, vague breach language, or claims of 'industry -standard' security without operational detail."

Meghna Manoj Nair 2026 (3)¶

Meghna Manoj Nair, Rachel Greenstadt. (2026). "We'll Fix It Later": Education, AI, and the Deferral of Student Privacy in EdTech. New York University. https://arxiv.org/abs/2609.28137

q2 · i? · qualitative · r3

Audit finding from the Results section: C1 scored 0% at zero, 21% at one, and 79% at two, mean 1.81 of 2, so platforms describe collection better than subsequent governance of that data.

"Data collection disclosure (C1) was strongest, while downstream governance was weaker."

Discussion¶

Related Claims¶

• EdTech organizations delegate privacy responsibility to cloud providers, policy documents, and downstream institutions, diffusing accountability — related
• K-12 platforms scored higher on children's consent where regulation is explicit, but the advantage did not extend to AI disclosure or accountability — related
• Detailed data disclosure statements, in either text or nutrition-label format, increase university students' confidence that specific data points are collected compared to a generic disclosure — related
• Detailed disclosure does not make students more conservative about who should have access to their educational data — related
• Policy packages fall into four archetypes defined by which design element is missing — related
• University AI policies overwhelmingly use weaker, suggestive language rather than strong directive language — related
• Safeguards in university AI policies are sparse and unevenly covered — related