2026-10-29 · MIT Tech Review
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…
原文链接：https://www.technologyreview.com/2026/10/29/1147824/monitor-ai-agents-actions-real-time/
查看更多资讯请见 AI 行业资讯。