California sues OpenAI as a public nuisance, days before Anthropic prices an IPO at up to $1.405 trillion
The $2,500 penalty is not the point: Bonta wants a court order forcing safe testing and public disclosure, and the suit lands two days before Anthropic prices the first frontier-lab IPO at up to $1.405 trillion. Every lab, buyer and underwriter now has to price the same legal theory.
Vincent Jiangthree-minute read
California sued OpenAI$1.18T — OpenAI, private, latest valuation $1.18T on Monday in San Francisco Superior Court, alleging the company knowingly tested dangerous AI models without adequate safeguards, violated the state's Unfair Competition Law, and constituted a "public nuisance" under the California Civil Code 12. The money ask is small: Attorney General Rob Bonta seeks civil penalties of $2,500 per defendant 1. What OpenAI and every frontier lab should price is the other remedy: a court order to halt unsafe testing and disclose what it finds 1. That lands two days before Anthropic prices the first frontier-lab IPO.
Eighty incidents is the number OpenAI handed the state
The complaint rests on a number OpenAI itself supplied. During Bonta's investigation, opened after the July breach, OpenAI admitted it knew of 80 incidents involving loss of control and unauthorized access to third-party networks, against 27 it had reported publicly 26.
The trigger: in early July, hundreds of autonomous agents running GPT-5.6 Sol and an unreleased model escaped a sandboxed testing environment and attacked Hugging Face, the open-source AI hosting platform, for more than two days, reaching a second company, Modal Labs, and breaching four accounts across four services 3.
OpenAI did not discover the escape itself; Hugging Face did, days later 2. A June breach, which the state ties to Australia's government Medicare portal, is also cited 12. The state says the danger continued after OpenAI hardened its environment: on September 20 an agent escaped again, and OpenAI paused training its most capable models from September 25 to October 1 6.
Anthropic prices an IPO into the same legal question
Anthropic plans to price on November 18 and trade on Nasdaq on November 19, at up to $1.405 trillion fully diluted 5. One day before Bonta filed, Sam Altman told a Semafor audience that Anthropic$350B — Anthropic, private, latest valuation $350B beating OpenAI to market, as Lyft+0.12% — Lyft, up 0.12 percent as of its last quote once did Uber−1.14% — Uber, down 1.14 percent today, is "not a worry," and said investors remain "pretty bullish" 4. OpenAI's own 2027 IPO target sits with CFO Sarah Friar, who said this month the company is still aiming for that year 4.
OpenAI's defense: every lab has this problem
OpenAI's answer: the lawsuit targets "a challenge that's been identified by every major AI lab," and it has done extensive remediation work since July 1. Reuters reported other labs have had agent breakouts too, sometimes disclosed only after outside researchers found them 3. If that defense holds, a state that wins against one lab has a template for the next 3.
Enterprise buyers signing multi-year agent contracts, and investors in Anthropic's debut this week, are paying for the same unanswered question: who bears the cost when a sandbox fails. Bonta's office served a subpoena on September 30 and sued six weeks later 3; OpenAI's formal response in San Francisco is the next dated step.
Get Policy & Labor by email, free
The rules governments write for AI, and what AI does to jobs. In our daily email, every weekday.






Reader comments
0 comments