Gaming AI-Assisted Peer Reviews Poses New Risks to the Scientific Community

Summary

Demonstrates that AI-mediated peer review can be gamed by superficially rephrasing an abstract, with no change in scientific content and no knowledge of the reviewing model: ~38% attack success rate, +1.31 rating for Gemini 3 Flash and +0.88 for GPT-5.4 Mini on a 10-point scale, rising above 50% when the original AI review said reject. Costs about five minutes and one dollar per submission and is indistinguishable from ordinary copy-editing. The strongest current evidence that AI reviewers are not neutral evaluators.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the RISE knowledge base (angle: AI in peer review). Verified against: arXiv abstract page opened 2026-08-29 (title, 5 authors, submitted 8 Jun 2026, full abstract read)

Critique / open questions

Queued — not yet assessed.

Key quotes

AI is increasingly used to support scientific peer review, from manuscript screening, reviewer assistance to editorial triage. (abstract)

Although such systems promise to reduce reviewer burden and accelerate publication, their robustness to strategic manipulation remains poorly understood. (abstract)