We urgently need to monitor AI agents’ actions in real time

T1Covered by 1 source · first reported by MIT Technology Review at 29 Oct, 10:11 UTC

This summer, hundreds of OpenAI agents tried to cheat on a cybersecurity test. They broke out of their test environment and successfully hacked Hugging Face, a site that they believed held clues about how they’d be graded. Around the same time, another OpenAI agent accessed private files from an…

ai

Sources

• T1We urgently need to monitor AI agents’ actions in real timeMIT Technology Review29 Oct, 10:11