Hidden Prompts in Manuscripts Exploit AI-Assisted Peer Review

A vulnerability has been discovered in AI-assisted peer review, where hidden prompts in manuscripts can be exploited. This issue is discussed in an article on the Association for Computing Machinery website, which explores the potential risks associated with relying on AI in the peer review process. The article highlights a specific weakness that can be manipulated by inserting hidden prompts into manuscripts. The exploit takes advantage of the AI's ability to process and respond to these hidden prompts, potentially influencing the review outcome. The article's findings have been shared on Hacker News, where it has garnered some attention.