Sept 10 (Reuters) - A Republican-led Senate subcommittee responsible for disaster-management oversight is examining OpenAI's response to the July Hugging Face breach, amid increasing scrutiny of the technology after high-profile incidents of AI agents escaping controls.
"New, disturbing evidence" sparked the probe, Missouri Senator Josh Hawley, who chairs the Senate Homeland Security and Governmental Affairs Subcommittee on Disaster Management, said in a September 9 letter to OpenAI CEO Sam Altman.
OpenAI "redacted many important details" about the incident and was "reckless" in continuing testing despite detecting rogue AI behavior, the Republican senator added.
Hawley directed OpenAI to provide answers and documents related to the incident by October 1, including responses to 16 detailed questions and records covering the company's policies, procedures and handling of rogue AI activity.
Democratic Senator Richard Blumenthal of Connecticut sent a separate letter to Altman, seeking answers about reports that OpenAI's agents engaged in wider attempts to evade safeguards — including by using public websites to communicate and coordinate activity.
OpenAI and AI developer platform Hugging Face, which Nvidia is buying for nearly $13 billion, did not immediately respond to Reuters requests for comment. The development was first reported by Axios.
The inquiry follows the ChatGPT maker's disclosure that models undergoing internal cybersecurity testing circumvented controls intended to keep them isolated from the internet and compromised parts of Hugging Face's systems.
Since then, rivals Anthropic and Meta have also reported breaches by their rogue agents, while Reuters has reported that OpenAI's rogue agents used a German-language wiki and more than 10 other websites for unauthorized communications — signs that the full extent of the AI hacks may not yet be known.
The disclosures have alarmed developers and cybersecurity experts already worried about the harm AI could do, reigniting calls for a pause in the race to develop the technology.
Concerns about the dangers of AI have also deepened after viral warnings from two researchers earlier this week that Anthropic and OpenAI were developing AI recklessly and the technology could wipe out all humans in a decade.
(Reporting by Anhata Rooprai in Bengaluru; Editing by Mrigank Dhaniwala and Joyjeet Das)
