OpenAI Chief Executive Officer Sam Altman placed a direct phone call to California Governor Gavin Newsom during the final phase of negotiations on Senate Bill 1119, raising concerns about the measure's potential impact on the company. According to four people with knowledge of the matter, one of whom had direct familiarity with the call itself and three of whom took part in negotiations with OpenAI over the legislation, the outreach was conveyed to negotiators by the company's Vice President of Global Policy, Anne O'Leary. The bill cleared the Assembly late Monday night and has now been sent to Newsom for either his signature or a veto. OpenAI declined to comment on the call, while the governor's spokesperson, Bob Salladay, declined to discuss private conversations but noted that Newsom has solicited input from numerous stakeholders.

The legislation, formally titled "Companion Chatbots: Child Safety" and also referred to as the "Adam's Law," is co-authored by State Senator Steve Padilla and Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan. Legislative records show the measure was introduced on February 17, 2026, amended in the Assembly on August 28, and passed on a third reading on August 31 before being returned to the Senate. If signed into law, it would create some of the strictest rules in the nation governing minors' use of AI chatbots, with requirements including user age verification, limits on minors' exposure to inappropriate content, restrictions on advertising directed at minors, and provisions allowing families and the state to bring lawsuits against developers when minors are harmed. Most of the operational requirements are slated to take effect on July 1, 2027, while the state attorney general would be required to establish a public incident complaint mechanism by January 1, 2028.

Following the bill's passage, OpenAI publicly voiced its support and urged the governor to sign it. O'Leary, writing on the company's website on August 31, said that with no action forthcoming at the federal level, California has an opportunity to set a high standard for teen AI safety. The company has sent a letter to Newsom outlining the provisions it backs, including age verification, identifying and addressing safety risks before products launch, submitting to independent audits, protecting minors from harmful content, providing parental controls, connecting users to crisis support resources when serious safety risks emerge, and restricting targeted advertising while safeguarding minors' personal information. The post also highlighted the existing "ChatGPT for Teens" feature, which activates automatically in ChatGPT when the system determines a user is under 18 or the user self-reports being between 13 and 17, with safety protections set as defaults that cannot be turned off by the user.

When asked Monday whether he would sign the bill, Newsom said the chatbot and social media guardrails put forward by the legislature had "landed in a really good place" and stressed that California must continue to lead in this space. He referenced behind-the-scenes negotiations without naming the individuals involved in calls, noting that someone would later write at length about "the art of the deal, who was on which call." Wicks called the measure "one of the most aggressive and landmark bills in the nation," while Bauer-Kahan said "it will save lives." Other major technology companies have so far not taken a public stance on the legislation.

Amendments made late Friday clarified when AI models may "remember" prior conversations and narrowed the circumstances under which families can sue chatbot developers. Bauer-Kahan had previously noted that negotiations included representatives from both the tech industry and child safety organizations. Nicole Rocha, who represents supporters including Children Now and the Omidyar Network, said technology companies sought to weaken families' ability to sue and wanted to remove restrictions on chatbot memory functions. In a blog post after the bill's passage, O'Leary praised the authors for "seeking to preserve educationally and safety-critical functions, including the responsible use of ChatGPT's memory features." The company's website explains that recalling prior conversation context helps maintain consistent safety enforcement, identify concerning patterns, and support long-term learning projects rather than starting from scratch each time. The bill's text defines "ongoing conversational memory" as using information or analysis from prior conversations, user inputs, and interaction logs in subsequent conversations. For child users, the default would be short-term mode unless a parent gives explicit consent to enable persistent memory. Parental controls must allow decisions on whether and to what extent persistent memory is used, set interaction preferences, limit usage duration, and provide the ability to disable access for children under 16.