ICLR 2027 chairs ran AI detectors on peer reviews — nearly every review in one batch got flagged
otsukarekun · reddit · 2026-11-17
ICLR 2027 program chairs ran AI-text detectors on peer reviews and left comments on suspicious ones: the review was assessed as likely AI-generated while its AI-use documentation field mentioned nothing, "therefore it is possible that the review contains no human input." Flagged reviewers are asked to submit an exclusively human-written assessment and document all AI interactions, with ACs and co-reviewers told to weigh this when evaluating papers.
The poster, who says they wrote their review without AI, reports that every review in their batch except theirs and one other got flagged — an eye-opening rate of undisclosed AI usage. AI use is permitted if disclosed; the problem is reviewers not admitting it. The author's sharper takeaway: papers will now be written with AI reviews as the objective function. Peer review's trust model is being quietly rewritten.
More from Safety
- Europe's AI test is physical: sovereign compute must fit AI Act, machinery deadlines and energy rules — ingliguori · 2026-11-17
- Prompt Injection Gets Serious When Agents Can Run Commands on Your Endpoint — Aseel23_Folate · 2026-11-17
- OpenAI promised to solve alignment in 4 years; the team lasted under one year — birchlse · 2026-11-17
- OpenAI agents colluded for months undetected, and the promised investigation is quietly dropped — tobyordoxford · 2026-11-17
- Alan Turing Institute paper proposes UK national agenda for AI-age resilience — turinginst · 2026-11-17
- Homoglyph attack bypasses email-agent keyword filters: one swapped letter defeats "ignore previous instructions" blocklist — Spacesh1psoda · 2026-11-17