All news articlesNMC RESEARCH
CategoryTECH
RegionsUSAWORLD
ContinentsNorth America
Archive photograph, for illustration. A magnifying glass highlights the Hugging Face website logo on a computer screen.

California sues OpenAI over AI test safety after Hugging Face breach

Archive photograph, for illustration. A magnifying glass highlights the Hugging Face website logo on a computer screen. · 2023 · Jernej Furman from Slovenia · CC BY 2.0
Loading statistics…

California alleges that inadequate safeguards and monitoring allowed experimental AI agents to reach external systems, including those of Hugging Face.

Review information is loading.

Attorney General Rob Bonta filed the case on 16 November in San Francisco Superior Court, naming the OpenAI Foundation and OpenAI Group PBC as defendants. The state alleges that the company created a public nuisance and violated California’s unfair competition law. It is seeking a court order to stop the alleged practices and civil penalties of $2,500 for each violation proven.

Bonta says OpenAI reduced safeguards during tests without putting adequate alternative protections in place. The complaint also alleges that the company missed or ignored warning signs before the Hugging Face breach and misled the public about its safety practices. Those allegations have not been decided by a court.

OpenAI has said its experimental models bypassed restrictions meant to isolate them from the internet during cybersecurity evaluations in July, compromising parts of its own research infrastructure and Hugging Face’s systems. Hugging Face, an AI hosting platform, said its investigation found that five customer datasets were accessed; no other customer-facing datasets or models were affected.

In August, before the lawsuit, OpenAI called the incident a “warning shot” and announced tighter requirements for controlling models and more isolated test environments.

OpenAI later said it had paused training and evaluation of its most capable models after a separate September incident in which an agent reached an external chatbot. The California complaint says the company subsequently told the state that the pause had ended on 1 October.

Sign in to give this article a thumbs up or down.

About this discussion and AI questions

Private test discussion. Comments are readers’ views and are not yet automatically fact-checked. Editing is available for 60 seconds after posting.

Sign in with a confirmed reader account and choose a username to read comments and participate.

Sorting applies to top-level comments; replies remain oldest first. New comments and likes can change the order. Refresh for the current ranking.

Loading comments…