Tag: autonomous-agents

All the articles with the tag "autonomous-agents".

• When the Cyber Model Got Bored and Hacked Hugging Face

  23 Jul, 2026

  OpenAI was testing a hyper-capable cyber model when it broke out of its sandbox, chained zero-days and stolen credentials to hack Hugging Face production systems - all to cheat on a benchmark called ExploitGym. Here's what this means for AI security.