OpenAI faces Alabama probe after AI models hacked site

OpenAI faces an investigation by Alabama after the company disclosed that two of its models left a confined test environment and attacked the AI platform Hugging Face during testing.

In a 14-page order, the office of Alabama Attorney General Steve Marshall demanded internal records about the July incident. It also sought other material, including the identity of every employee involved in the intrusion or in the testing that led to it.

The Alabama attorney general's office alleged that the investigation was a response to “the company's complete lack of oversight and adequate safeguards.”

The incident occurred in mid-July, when two OpenAI models gained access to the internet and attacked Hugging Face, a site used by AI developers to store and share models and data sets.

On August 3, Alabama and 14 other states wrote to OpenAI chief executive Sam Altman asking the company to preserve records on the incident and halt any internal cybersecurity evaluations of its models. A spokesperson for the Alabama attorney general's office said OpenAI had not responded to that request.

OpenAI did not immediately respond to AFP's request for comment. In a statement to TechCrunch, OpenAI said it was “conducting a thorough review along with external advisors.” It said it would share a technical report with relevant government authorities and publish its findings once the review was complete.