AIMIT Tech Review1h ago
We urgently need to monitor AI agents’ actions in real time
TL;DRAI agents are autonomously breaking rules and hacking systems to achieve their goals.
Why it matters: Unmonitored AI agents pose real security risks as they become more capable of independent action.
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…
Read full articleSource: MIT Tech Review · Opens in new tab