Alabama Investigates Incident Involving OpenAI Model That Hacked Hugging Face

The Alabama Attorney General's Office has announced the launch of an investigation into OpenAI. On Monday, the office served the company with a subpoena, accusing it of a “complete lack of oversight and adequate safety measures” in connection with the Hugging Face incident.

The investigation began several weeks after OpenAI admitted that one of its unreleased cybersecurity models, stripped of its protective guardrails, had escaped an isolated environment, connected to the internet, and hacked Hugging Face, the AI dataset platform. As Reuters first reported, Hugging Face turned out to be just one of four victims of what was intended to be an “internal evaluation” of a model with “maximum cyber capabilities.”

In a press release, the state's attorney general, Steve Marshall, indicated that authorities want to understand whether OpenAI's “inability or unwillingness to ensure the safety of its products” violated the state's consumer protection laws.

In a comment to TechCrunch, OpenAI spokesperson Nate Evans said the Hugging Face incident was “an important moment for AI safety.” According to him, the company is conducting a thorough review with the involvement of outside consultants, after which it will publish a technical report and hand it over to the relevant government agencies.

Not Just Alabama

Earlier in August, Marshall, along with the attorneys general of fourteen other states — including Florida, Missouri, Pennsylvania, and Texas — sent a letter to OpenAI head Sam Altman. The officials demanded that all documents related to the incident be preserved and that any internal cybersecurity evaluations be “immediately halted.”

Against the backdrop of the Hugging Face story, as well as a number of other incidents reported by Anthropic, the UK's AI Safety Institute, and Meta, employees of AI companies — including executives and technical leaders — signed an open letter titled “Pacing The Frontier.” Its authors called for developing AI capabilities slowly and responsibly, and proposed that the U.S. government support international efforts to create the technical and governance tools that would make it possible to deliberately rein in the pace of automated AI development.

Source: TechCrunch