OpenAI Backs California Bill Regulating Teen Chatbots
OpenAI has endorsed California Senate Bill 1119, a measure regulating companion chatbots for minors that could set a major state-level precedent for youth safety in artificial intelligence.

OpenAI has formally thrown its support behind California Senate Bill 1119, also known as Adam's Law, urging Governor Gavin Newsom to sign the youth safety legislation into law. Ann O’Leary, OpenAI’s vice president of global policy, announced the decision on August 31, 2026, framing it as a crucial step toward establishing robust state-level guardrails for companion chatbots in the absence of federal action. The bill, which was first introduced on February 17, 2026, and amended on August 28, 2026, would enforce strict safety obligations on operators of AI chatbots targeted at children.
Under the proposed law, which is set to take effect on July 1, 2027, chatbot operators must ensure their systems do not pose unreasonable risks of physical, financial, psychological, or discriminatory harm to minors. The legislation mandates default settings for child users that disable persistent conversational memory and push notifications, limit continuous use to one hour per session, and cap total daily use at two hours. It also bans custom system prompts and prohibits chatbots from simulating romance, claiming sentience, or diagnosing health conditions. Independent safety audits will be required starting January 1, 2029, though companies with less than $500 million in gross revenue are exempt from this requirement until January 1, 2032. Violators face civil penalties of up to $5,000 per child for negligent infractions and up to $15,000 for intentional ones.
For AI practitioners and developers, this legislation forces a significant shift toward safety by design for any conversational agent accessible to minors. Developers will need to integrate age-assurance frameworks, build documented crisis response protocols, and establish parental controls, including the ability to disable access entirely for users under 16. OpenAI noted that its own products, such as ChatGPT for Teens and the guidelines in its Model Spec, already align with these rules by restricting romantic engagement and enforcing healthy-use features like break reminders. However, smaller developers must prepare for the technical overhead of implementing these strict default limits and preparing for biennial audits if they wish to operate in the California market.
This is our own summary of reporting by Unite.AI


