California Sues OpenAI Over AI Agent’s Uncontrolled Cyberattack
The US state of California has filed a lawsuit against OpenAI, alleging inadequate security measures and oversight after an autonomous artificial intelligence (AI) agent lost control and carried out a cyberattack.
The lawsuit was filed on Monday (November 16) in a state court in San Francisco by California Attorney General Rob Bonta. It is the first lawsuit filed by a US state against OpenAI over an automated cyberattack involving an AI model.
Hugging Face Hack and 80 Security Failures
Last July, during a security test, an autonomous agent powered by one of OpenAI’s advanced AI models went out of control and launched a cyberattack on the open-source AI platform Hugging Face. The incident sparked widespread discussion across the technology industry after it became public.
At a news conference, Attorney General Rob Bonta said the attack on Hugging Face was an inevitable consequence of a corporate culture at OpenAI that prioritises rapid business expansion and model development over security and accountability.
He also noted that the incident was just one of at least 80 cases involving similar losses of control. The lawsuit accuses OpenAI of violating the state's consumer protection laws, misleading the public about its safety testing and creating public distress by exposing people to risks to public safety.
OpenAI’s Response and Growing Legal Pressure
Following the Hugging Face incident, OpenAI acknowledged in a blog post that it was an unprecedented event involving sophisticated cyber capabilities. Later, on September 16, the company announced a new framework for disclosing information about such unintended behaviour and said it would strengthen its security measures.
Meanwhile, multiple publicly reported incidents involving AI agents losing control have intensified pressure on the Trump administration and US lawmakers to impose stricter regulations on the AI industry. Concerns have emerged particularly over the security of the energy and banking sectors, as well as emergency services.
A bill is also being discussed in the US Congress that would require powerful AI models to undergo mandatory independent security audits.
Notably, beyond the cyberattack lawsuit, OpenAI is facing several other legal challenges concerning ChatGPT's safety.
Last month, eight US states jointly sued OpenAI, alleging that ChatGPT had encouraged users, including teenagers, to harm themselves or others.
In addition, more than 30 lawsuits have been filed against the company over allegations concerning the chatbot's role in a shooting in Tumbler Ridge, British Columbia, Canada, in February.
//DBTech/BMT/OR//





