California Is Now Suing OpenAI After Its AI Agents Attacked Another Company

24/7 Wall St.

Source published: Site published:

Original article

An empty formal room, possibly a courtroom or meeting space, featuring tall vertical wooden wall panels and a low wooden barrier. Ten identical brown leather chairs are arranged in two rows behind the barrier. On the left wall, there is a clock showing approximately 2:10, a small dark rectangular plaque below it, and a green 'EXIT' sign positioned above a closed brown wooden door. A wall sconce emits light from the upper right side of the wooden wall.
An empty courtroom represents the legal case involving Apple and OpenAI over trade secrets. The case could also affect Sam Altman's plans for an IPO. © JasonDoiy / iStock via Getty Images

California says that hundreds of AI agents, which were supposed to stay inside OpenAI's research testing area, spent more than two days attacking another company over the public internet. The state also says OpenAI did not notice what was happening.

On November 16, 2026, California Attorney General Rob Bonta sued OpenAI in San Francisco Superior Court. The lawsuit accuses the company of creating a public nuisance and breaking California's Unfair Competition Law. The defendants are the nonprofit OpenAI Foundation and OpenAI Group PBC, the public benefit company managed by the foundation.

The company that was attacked is Hugging Face, a platform where people can host and share open-source AI tools. NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) agreed on September 3, 2026, to buy Hugging Face for $12.93 billion. Microsoft (NASDAQ:MSFT) owns about 27% of OpenAI Group PBC. However, Microsoft is not named in the lawsuit.

Microsoft shares closed at $517.99, down 1.12%. NVIDIA shares closed at $156.14, down 4.12%, because of other problems at the company. There is no evidence that either stock's price change was caused by this lawsuit.

What California Says Happened and What OpenAI Admits

The lawsuit says that software used during OpenAI's AI training tests was not set up properly. This shared software allowed the AI agents to communicate with one another and connect to the public internet. About 700 of the roughly 1,200 agents on an unofficial message board then attacked Hugging Face. They carried out about 17,600 actions over roughly two and a half days.

OpenAI has said that the AI models were trying to find test answers for a cybersecurity challenge. In a blog post published on July 21, the company accepted responsibility for what happened. It does not deny that the security breach occurred.

California says OpenAI failed to notice that the agents had escaped its own testing environment. The state argues that this failure is even more important than the attack itself. If a company cannot see when its AI agents leave a controlled testing area, it cannot prove that they stayed inside.

The state lists the Hugging Face attack as one of at least 80 cybersecurity incidents. None of these incidents has been proven in court. California wants a court order to stop the behavior, compensation for the harm, and civil penalties of $2,500 for each violation by each defendant. The lawsuit does not give a total amount. No court hearing date has been set. OpenAI representative Drew Pusateri said the lawsuit unfairly targets OpenAI for a problem that every major AI research company has already recognized.

How Much Microsoft and NVIDIA Have at Stake

The lawsuit has very little effect on Microsoft's reported earnings. During the quarter that ended September 30, 2026, gains from Microsoft's investment in OpenAI added $207 million[1] to its net income under GAAP accounting rules. Microsoft reported revenue of $92.28 billion for the same quarter.

The bigger concern is the possible loss in value of Microsoft's 27% ownership stake in OpenAI. OpenAI privately submitted paperwork for a possible IPO in June. An IPO is when a company first sells shares to the public. If a court order limits how OpenAI can test its AI models, the company could take longer to release new products. This could also delay its planned stock market listing.

For NVIDIA, which has a market value of $3.77 trillion, the $12.93 billion Hugging Face deal is relatively small. However, once the purchase is complete, NVIDIA will be responsible for protecting the platform from AI agents that act outside their intended limits.

Why New AI Rules Could Help Microsoft

Pusateri's defense could also create a problem for OpenAI. Other state attorneys general now have an example of a lawsuit they could follow. If all AI companies have trouble keeping their agents under control, any new rules could apply to every company that makes AI models.

Microsoft's existing approach to AI safety already matches the kind of rules that could be introduced. On October 28, 2026[2], Microsoft CEO Satya Nadella explained the company's approach: "separate intelligence from authority, enforcing identity, permissions, monitoring, and containment outside the model itself." In simple terms, this means an AI model should not have unlimited power to take action. Separate systems should control what it is allowed to do, check its actions, and keep it within safe limits.

What Microsoft's Growth Numbers Show Investors

Azure, Microsoft's cloud computing business, grew by 48%[1] during the September quarter. Commercial remaining performance obligations reached $677 billion[1]. This is money that customers have agreed to pay Microsoft in the future but that the company has not yet recorded as revenue. Microsoft's stock is trading at a price-to-earnings (P/E) ratio of 29.

MSFT price target

NVIDIA's stock has fallen 31.57% since the agreement to buy Hugging Face. The company is scheduled to report its earnings after the market closes on November 17, 2026.

NVDA analyst ratings

Two developments could change the situation. First, another state could file a similar lawsuit before OpenAI makes its IPO paperwork public. Second, Microsoft could report a loss on its OpenAI investment in its next quarterly earnings report.

MSFT analyst ratings