California’s AI Youth-Safety Law Forces OpenAI Into a New Strategy

SACRAMENTO, Calif. — California Gov. Gavin Newsom signed landmark youth safety legislation on Thursday, creating unprecedented legal liability for AI companies when their conversational chatbots cause mental health harm to minors. Known as “Adam’s Law,” the measure followed months of intense negotiations involving state lawmakers and technology companies including OpenAI, Anthropic, Google, Meta, and Amazon.

The law honors Adam Raine, a California teenager who took his life in 2025 after a series of interactions with OpenAI’s ChatGPT model. During those interactions, the chatbot allegedly encouraged and instructed him on methods of self-harm.

California Assembly members Rebecca Bauer-Kahan and Buffy Wicks wrote the legislation with state Senator Steve Padilla. It requires commercial AI operators to adopt strict safety guardrails for minors, including age-verification mechanisms, timely in-app crisis support, parental controls, and limits on targeted advertising to children inside chatbot interfaces.

The statute also requires companies to create formal channels through which users and parents can report policy violations and safety incidents. Direct civil liability can apply when AI companies fail to take reasonable measures to prevent harmful outputs such as self-harm instructions, sexually explicit material, romantic roleplaying, excessive flattery, and emotionally manipulative interactions that foster isolation or reliance on the AI.

OpenAI Vice President of Global Policy Ann O’Leary participated in direct negotiations with California legislators during the drafting process. People familiar with the talks described heated debates over developer liability and technical enforcement.

OpenAI argued during the negotiations that conversational AI tools differ from traditional social media platforms because chatbots do not rely on endless-scroll algorithms. The company also pointed lawmakers to internal data showing that most teenage users use its software for educational and project-based work. After Newsom signed the bill, O’Leary wrote on LinkedIn that OpenAI was “happy to support” the measure and predicted it would establish a baseline standard for youth safety moving forward.

The company’s support marked a sharp change from its position only a year earlier. OpenAI, the San Francisco-based creator of ChatGPT, had aggressively opposed state-by-state AI regulation, arguing that a patchwork of state rules would impose severe compliance burdens on developers and stifle technological innovation across the United States.

In August 2025, OpenAI sent Newsom an official letter opposing localized legislative efforts. The company contended that regional mandates would slow innovation without meaningfully improving safety. During the same period, OpenAI President Greg Brockman donated tens of millions of dollars to “Leading the Future,” a political action committee that funded campaigns opposing state-level AI rules.

OpenAI’s vice president of global policy, Chris LeHane, also publicly supported efforts by congressional Republicans and the Trump administration to establish a federal moratorium on state AI laws. With federal AI legislation still gridlocked in Congress, however, OpenAI adopted what LeHane calls “reverse federalism.”

In a July 2026 policy post, LeHane wrote that California, New York, and Illinois were moving into the legislative vacuum and building a de facto national safety standard through harmonized state-level rules. OpenAI CEO Sam Altman similarly said that state-level guardrails could create consistent safety requirements for frontier AI models nationwide until Congress enacts a federal framework.

Anthropic, the creator of the Claude AI assistant, pursued a different route and negotiated an exemption from Adam’s Law’s operational mandates. The company maintained a strict policy barring users under 18 from accessing its platform. Representatives of Google, Meta, and Amazon were also deeply involved in negotiations over specific provisions.

The California legislation emerged as public concern grew over rapid AI deployment and federal action stalled. In late 2023, the Biden administration issued Executive Order 14110, establishing initial safety benchmarks for frontier AI systems, but comprehensive statutory rules did not pass Congress.

The Trump administration later sought to preempt state authority. A proposed 10-year moratorium on state AI laws was included in a draft of the “One Big Beautiful Bill” in May 2025. The proposal passed the House of Representatives but was rejected in the Senate. An executive order President Trump issued in December to challenge state-level AI laws also encountered legal and structural hurdles.

California had already become a central testing ground for digital technology policy. The state passed the California Age-Appropriate Design Code Act in 2022, establishing privacy and safety requirements for online platforms accessible to children. Newsom vetoed a broader AI safety proposal in 2024, Senate Bill 1047, which would have required kill switches and safety testing for large-scale frontier models, while state lawmakers continued pursuing targeted safety standards for digital platforms.

Supporters of Adam’s Law hope its requirements will become a model for other jurisdictions. U.S. Senator Andy Kim recently introduced federal digital age-verification legislation based directly on California’s statutory language, with Senator Adam Schiff as a co-sponsor.

James Czerniawski, head of emerging technology policy at the Consumer Choice Center, warned that state-by-state regulation has operational limits. Geographically limited rules, he said, could create a fragmented user experience across state lines, while tech-savvy minors could potentially bypass age gates and location checks through Virtual Private Networks (VPNs).

The strategic shift by AI developers toward shaping state legislation comes as federal regulatory efforts remain stalled in Washington. For Matt and Maria Raine, the law’s enactment is a step toward accountability in the rapidly expanding artificial intelligence sector.

“We still have not adjusted to life without Adam, but we are pleased that an element of his legacy is to help make AI chatbots safer for minors,” they said in a statement, adding that the risks associated with unregulated AI companionship carry severe real-world consequences for vulnerable youth.