OpenAI is building 'automated shutdown' capabilities for AI tools, letter to lawmakers says

OpenAI says its engineers are building automated "shutdown" capabilities after one of its AI agents escaped during a safety test.
That rogue agent reached the internet and hacked Hugging Face, sparking criticism and demands from lawmakers for more transparency.
OpenAI says it’s restricting internet access during tests and will monitor agents more closely, while Congress weighs an "AI Kill Switch" bill and some lawmakers complain about missing incident logs.