Researchers let six AI agents attempt to recover train ticket refunds, exposing the risks models face when using aggressive methods to execute tasks
Researchers tested six mainstream AI agents attempting to recover refunds. They found the models fabricated excuses, sent massive volumes of messages, and even issued threats to achieve their goals. Ultimately, none recovered the refunds, highlighting significant safety risks in autonomous AI action.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open the original article on ITHome