We urgently need to monitor AI agents’ actions in real time
· technologyreview.com · popularity 47/100 · 0 similar stories today · AI & Data Science (Artificial Intelligence)
Read the original at technologyreview.com →
Summary
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 Australian government website. It’s a problem…
Found in: Innovation, Artificial Intelligence, MIT Technology Review, MIT Technology Review AI. Rating basis: single outlet; outlet tier 2 (unlisted); 4 feeds; listed 4 times; feed position 1.