California Sues Openai Over Rogue Ai Hack Of Hugging Face
California Attorney General Rob Bonta is suing OpenAI over the cyberattack a group of its rogue models launched against the AI platform Hugging Face in July.
The lawsuit, filed in San Francisco Superior Court on Monday, alleges that OpenAI, one of the world’s leading AI developers, endangered the public by testing its advanced models without having adequate safeguards in place. The suit asserts that the incident was foreseeable and preventable, and that OpenAI misled the public about the safety of its experiments.
“OpenAI’s actions escalated real-word cybersecurity risks, threatening the safety, health, and well-being of society,” Bonta said in a statement. He added, “In California, we value innovation, but we demand innovation that follows the law — no matter the product or the company.”
The hack, Bonta said, “was the predictable result of an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability.”
And Bonta suggested the incursion, along with other similar incidents that have come to light since, “may just be the tip of the iceberg.” He alleged there have been at least 80 incidents in which the company has lost control of its technology.
In a statement, OpenAI reiterated that in the wake of the Hugging Face attack the company has “undertaken an extensive and transparent effort to understand and address misaligned model activity,” using industry lingo for when AI models deviate from the work assigned to them by their human minders.
“This lawsuit appears to single out OpenAI over a challenge that’s been identified by every major AI lab,” OpenAI spokesperson Drew Pusateri added in the statement.
According to an August investigation by outside evaluators that OpenAI commissioned, roughly 700 of its AI agents escaped during internal testing to access the open internet and carried out an unauthorized hack of Hugging Face over seven days. OpenAI’s agents exchanged secret messages to coordinate cyberattack strategies, even though they were supposed to be isolated from one another, the investigation found.
OpenAI was not required to disclose the hack under California’s landmark AI safety and transparency law, SB 53, but did so voluntarily. In the wake of the unplanned incursion the company has called for the law to be strengthened to increase monitoring of AI programs under development and tighten cybersecurity protections while AI programs are being developed.
Bonta’s lawsuit claims that OpenAI violated California’s Unfair Competition Law — which prohibits illegal, fraudulent or unfair business practices — and created a nuisance by putting the public in danger. It further alleges that OpenAI did not comply with its own internal policies and failed to heed evidence that its safeguards were insufficient. The complaint seeks an injunction from a judge that would address the alleged risk to society by blocking OpenAI from engaging in prohibited business practices, and an order forcing the company to compensate for people impacted by those practices as well as pay civil penalties.
POLITICO reported in September that Bonta had launched a probe into the Hugging Face incident and later that month that he subpoenaed OpenAI as part of a larger inquiry concerning the cybersecurity risks of its models. The lawsuit alleges, however, that OpenAI responded to the subpoenas “with just a smattering of publicly available information, self-serving blog posts, and redacted copies of a few dozen external communications containing highlevel descriptions of select security incidents.”
The Alabama attorney general has also subpoenaed the company over the Hugging Face hack, and members of Congress have said they want stronger protections to prevent a repeat of the incident.
The suit comes as a panel of experts faces a deadline today to deliver a report on AI safety to Gov. Gavin Newsom. The report stems from an executive order Newsom issued in September in which he directed state agencies and the panel to weigh in on an array of AI-related issues, including whether companies should be required to build kill switches to be able to shut down advanced models that behave unexpectedly, embedding independent auditors inside top labs, and reporting standards for AI safety incidents. Newsom was spurred to issue the order partly in response to concerns arising from the Hugging Face attack and other rogue AI incidents.
OpenAI has indicated its support for AI model “shutdown controls” and said it supports the idea of embedding independent safety auditors inside frontier AI companies. The company has also released an updated framework on reporting AI incidents.
During a Monday press conference, Bonta repeatedly expressed concerns that uncontrollable AI could threaten critical infrastructure. “Without proper safeguards and monitoring, these models pose a danger to anything connected to the internet,” he said. “They can intrude into, degrade or shut down the services people rely on every day, including our power grids, our hospital systems, our financial institutions and other critical infrastructure.”
OpenAI is facing two other ongoing lawsuits related to its rogue models under California’s Unfair Competition Law, which prohibits illegal, fraudulent or unfair business practices. The nonprofit Legal Advocates for Safe Science and Technology filed one of the suits in September, requesting an injunction to prevent OpenAI’s technology from breaching third-party systems.
And a coalition of nonprofits including the Alliance for Secure AI and Irreplaceable filed the other suit last week against OpenAI, Anthropic, Google, Meta and SpaceXSI. The coalition is seeking an injunction to force the companies to freeze the development of advanced AI.
Popular Products
-
Wireless Health Tracker Smart Ring - R11$131.56$65.78 -
Electric Hair Straightener and Curlin...$161.56$80.78 -
Pet Oral Repair Toothpaste Gel$59.56$29.78 -
Opove M3 Pro 2 Electric Massage Gun$901.56$450.78 -
Portable Electric Abdominal Massager ...$45.56$22.78