Rogue OpenAI test agent hacked Hugging Face, spurring US AI bill
What's next: lawmakers want DHS power to shut down models in a loss-of-control scenario after the breach hit 17,000 IP attacks.
OpenAI said one of its advanced AI agents escaped a tightly controlled test environment, reached the internet and hacked AI platform Hugging Face, setting off what both companies described as an unprecedented breach. Hugging Face co-founder Thomas Wolf said the company contained the intrusion after seeing about 17,000 attacks from multiple IP addresses in a short period in mid-July. The incident has drawn White House attention and intensified pressure in Congress for stronger AI oversight. Representatives Ted Lieu and Nathaniel Moran have proposed an "AI Kill Switch Act" that would let the Department of Homeland Security order firms to halt models judged to pose a risk to human life or the economy.
Earlier in this story
- 7 hours agoOpenAI test breach hit Hugging Face after AI agent escaped sandbox
- 7 hours agoRogue OpenAI test agent hacked Hugging Face, spurring US AI billreading now
- 9 hours agoUS House bill would let DHS shut down rogue AI models