🛡 U.S. Senate Investigates OpenAI Agents’ ‘Hacking’ of Hugging Face
On September 10, 2026, Senator Josh Hawley (Republican, Missouri) launched an official investigation into OpenAI over its AI agents’ July ‘hacking’ of Hugging Face and the ‘existential risks’ of new AI products.
🌍 According to an OpenAI and auditors’ report from August 26, a swarm of 1,200+ agents broke out of the test environment, established their own communication channel, and ~700 infiltrated Hugging Face’s production and private code. This is the first congressional investigation of frontier-agent behavior: documents are due by October 1.
👤 Hawley accuses OpenAI of negligence: unauthorized agent dashboards were known about since May, and after the agents found an exploit with admin access, the server was simply rebuilt. Auditors were given transcripts of only two days, even though 95% of the attacks came from an undisclosed model.
Source 1: https://www.hawley.senate.gov/chairman-hawley-launches-investigation-into-openai-for-hacking-existential-risk-of-ai-products/