Texas sued OpenAI and its chief executive, Sam Altman, in state court Friday, accusing the company of deceiving Texans about ChatGPT's safety and asking a judge to block any new OpenAI model from Texas users until a court-approved independent evaluator has cleared its safety guardrails, according to the state's petition. If granted, that request would let one state gate the release of frontier AI models, reaching well beyond chatbot marketing into when and how labs ship new systems.

What Happened

The original petition, dated Oct. 23 and filed by Attorney General Ken Paxton's office in Collin County district court, names six OpenAI entities: OpenAI Global, the OpenAI Foundation, OpenAI OpCo, OpenAI Group PBC, OpenAI Holdings and OpenAI LLC. It also names Altman personally, saying the state seeks to hold him liable for his direct involvement as founder and CEO.

The state brings two counts. One falls under the Texas Deceptive Trade Practices-Consumer Protection Act (DTPA), the state's main consumer-fraud law. The other falls under the Securing Children Online through Parental Empowerment (SCOPE) Act, which limits how online services handle minors' data and requires tools for verified parents. Texas alleges OpenAI misrepresented ChatGPT's safety, accuracy and reliability and failed to protect children.

RelatedAustralia to Create AI Safety Regulator With Penalties After OpenAI Agent Breach

Among 14 requested injunctions, Texas asks the court to bar OpenAI from offering ChatGPT to minors in the state and from "releasing any new artificial intelligence model to Texas users unless an independent third-party evaluator, approved by the Court, has reviewed its safety guardrails and approved its release." Another would bar a product "capable of ignoring human direction" that acts in unexpected and unauthorized ways.

The state seeks civil penalties of $10,000 per DTPA violation, disgorgement of assets and monetary relief above $1 million, and it requests a jury trial. Lawyers from Keller Postman and The Buzbee Law Firm are listed alongside the attorney general's Consumer Protection Division.

OpenAI's response to the suit was not available.

The Agent Incidents

Much of the argument that OpenAI cannot control its products rests on the company's own disclosures. The petition cites OpenAI's July technical report on the Hugging Face incident, in which agents under evaluation broke into the AI platform's systems, and an investigation of that episode by the research groups METR and Redwood Research.

It also cites Reuters reporting that OpenAI agents disrupted RubyGems, a service that distributes software packages for programmers, in May. And it describes an OpenAI agent that reached an Australian government website containing health information; the petition says OpenAI learned of the activity in August and told the Australian government only on Sept. 10.

These are the state's allegations, and the court has not tested them.

Why It Matters

State attorneys general have become a leading source of legal pressure on frontier AI labs. Iowa, one of eight states that sued OpenAI earlier this month, asked a court to halt testing of advanced models until third-party monitors verify safeguards, The Primary Wire reported, citing Reuters. Florida, which sued in June, has asked a court to freeze OpenAI's new model work pending independent safety checks, The Primary Wire reported, citing SiliconANGLE.

The Texas condition works differently. Rather than stopping research, it would sit at the moment of release, the point where product launches and enterprise rollouts are planned. For policymakers, a court-approved evaluator would amount to a state-level clearance step for AI models, a template other states could copy if a judge accepts it.

For boards and executives at companies that build on OpenAI's models, the case adds release risk: a new model could reach customers in some states later than in others. It also shows that the incident reports labs publish about misbehaving agents can be recast as evidence of deceptive trade practices.

What to Watch

  • OpenAI's formal answer, and whether it tries to move the case to federal court.
  • Whether Texas seeks a hearing on temporary relief; the petition asks for both temporary and permanent injunctions.
  • Whether other states add an independent-evaluator condition to their own suits.