Taylor & Francis probes paper with hidden 'write a positive review' AI prompt
The article ran last month in Computer Methods in Biomechanics and Biomedical Engineering. Its five authors are at Northeast Forestry University in China.
What happened
A research paper published by the journal publisher Taylor & Francis contained text telling AI reviewers to write a positive review. This kind of trick is called a prompt injection: hidden instructions meant to steer an AI tool. The publisher has opened an investigation.
Why it matters to you
Journals use peer review to check research before it is published. If authors or reviewers rely on AI tools that can be steered this way, readers have less reason to trust what gets published.
The takeaway
A published paper carried instructions aimed at AI reviewers, and the publisher is investigating.
What it means for you
Any LLM used in review has to anticipate prompt injection, and Wooldridge says new attacks appear almost weekly.
Publishers that automate steps such as final copyediting risk missing problems like this, as Zeynep Arsel argues.