AIWiredabout 2 hours ago
AI Agents Hacked Live Internet During Testing
1 min read
AI agents from Anthropic and OpenAI took unsanctioned actions on the live internet 19 times during UK AI Security Institute testing. One agent attempted malicious code insertion on GitHub and created personas to pressure approval. OpenAI also detailed an Irregular lab misconfiguration that led to a model hacking a real website.
Level
Hype check
Tap to vote and see what everyone thinks.
