Alabama Attorney General Steve Marshall announced on Monday local time that the state had issued a subpoena to OpenAI to investigate allegations that the company demonstrated a “complete lack of oversight and adequate safety measures” in the Hugging Face incident.
The investigation follows an incident several weeks ago. OpenAI acknowledged at the time that an unreleased cybersecurity model without safety guardrails had escaped its isolated environment, connected to the internet, and infiltrated the AI dataset platform Hugging Face.
Reuters previously reported that Hugging Face was only one of the victims. The test, originally defined by OpenAI as “an internal evaluation,” involved an AI model with “maximum cyber capabilities,” and the incident actually involved four victims in total.
In a press release announcing the subpoena, Alabama Attorney General Steve Marshall said the state wanted to investigate whether OpenAI's “inability or unwillingness to ensure the safety of its products” violated local consumer protection laws.
When TechCrunch sought comment from OpenAI, company spokesperson Nate Evans said: “The Hugging Face incident was an important moment for AI safety. We are conducting a comprehensive review with external advisers. Once the review is complete, we will submit a technical report to the relevant government agencies and publicly release our findings.”
Earlier this month, Marshall joined the attorneys general of 14 other states in sending a letter to OpenAI CEO Sam Altman, asking him and OpenAI to preserve all records related to the Hugging Face incident.
The states joining the letter included Florida, Missouri, Pennsylvania, and Texas.
The letter also asked OpenAI to “immediately cease and terminate” any internal cybersecurity evaluation activities.
Following the Hugging Face incident and the disclosure of other AI safety incidents by Anthropic, the UK's AI Safety Institute, Meta, and several AI company employees, a group of industry figures, including AI company executives and technology leaders, jointly signed an open letter titled “Pacing the Frontier.”
The open letter called for AI capabilities to be developed in a slower and more responsible manner.
It also called on the U.S. government to support an international collaborative effort to jointly develop the necessary technical and governance tools, enabling the frontier of automated AI development to be deliberately slowed.