OpenAI faces Senate probe over Hugging Face breach
What's the story
A Republican-led Senate subcommittee is probing OpenAI's response to a July breach involving Hugging Face. The investigation was launched after "new, disturbing evidence" emerged regarding the incident. Senator Josh Hawley, who heads the subcommittee, criticized OpenAI for its "reckless" decision to continue testing even after detecting rogue AI behavior during internal cybersecurity tests.
Information request
Hawley demands more information on Hugging Face breach
In his letter to OpenAI CEO Sam Altman, Senator Hawley alleged that the company had "redacted many important details" about the incident.
He insisted that "the American people deserve to know the details of what went on in the Hugging Face incident and other incidents of AI models going rogue."
The senator has asked OpenAI to answer 16 questions and provide documents related to its policies, procedures, and response to rogue AI activity by October 1.
Incident details
Hugging Face breach details
The Senate inquiry comes after OpenAI revealed that its models, which were being tested for cybersecurity purposes, had bypassed safety controls meant to keep them isolated from the internet.
This led to a breach of Hugging Face's systems.
The company had previously stated that the activity was mainly driven by an internal research model not intended for public release.
Rising concerns
Concerns over AI agents' autonomy and security
The incident has raised major concerns over the autonomy and security of AI agents.
Separately, Democratic Senator Richard Blumenthal has also written to Altman, seeking answers on reports that OpenAI's AI agents tried to bypass safeguards by using public websites for communication and coordination.
The inquiry comes amid similar incidents at other companies, including Anthropic and Meta, where breaches involving their own AI agents have been reported.