California lawmakers have approved sweeping AI chatbot safety legislation aimed at protecting minors, sending the measure to Gov. Gavin Newsom less than a week after OpenAI chief executive Sam Altman made a reported attempt to lobby the governor’s office. The bill, SB 1119, cleared the state legislature on Monday and is now one signature away from becoming what supporters have described as a national reference point for how AI companies design conversational products for children.

Key facts

• SB 1119 passed the California Legislature on Monday and now goes to Gov. Newsom, who has not said whether he will sign it.
• OpenAI’s Sam Altman reportedly tried to reach Newsom last week while the bill was still being finalized, though the exact nature of any conversation remains unclear.
• The bill would require AI companies to determine user age, provide parental controls, restrict certain advertising to minors, and submit to independent audits.
• The measure was amended on Friday in ways that narrowed some civil-liability provisions, according to reports.
• OpenAI publicly supports the bill and has urged Newsom to sign it, pointing to its newer ChatGPT experience designed for teenagers.

A wide-ranging children’s safety measure

SB 1119, introduced by state Sen. Steve Padilla and Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, was designed to respond to a period of mounting alarm about the risks of conversational AI. The bill would treat chatbot services as products with a particular duty to protect the youngest users, rather than as neutral websites or platforms.

One of the central pillars is age assurance. Companies that build chatbots, including general-purpose assistants such as ChatGPT, would be required to determine a user’s age or at least reasonably estimate whether the user is under 18. That requirement matters because AI services often have no way of knowing who is on the other side of a conversation. Without an age checkpoint, a child can receive responses intended for an adult, including content that a psychiatrist or pediatrician would consider developmentally inappropriate.

The bill also places an emphasis on parental control. It would oblige companies to supply tools that allow parents to limit how much time their children spend inside a chatbot conversation and to shape when and how the AI product can be used. OpenAI has argued that many of these concepts are already part of its “ChatGPT for Teens” experience, which it launched on Aug. 18. In that configuration, the company says, users between 13 and 17 are automatically placed into a safer version of the product unless they have stated they are older.

Another significant feature of SB 1119 is its treatment of data and advertising. The measure would bar AI companies from serving advertisements to children when those ads are based on any personal information gathered from the child’s chatbot sessions. Selling data collected from a minor’s chat history would also become illegal. The bill does allow “age-appropriate contextual advertising,” but even those ads must be clearly labeled so a young user does not confuse them with a chatbot answer.

Lawmakers amended the measure Friday. The tech industry had been watching those last-minute changes closely, especially anything that could affect the likelihood of litigation. Reports say the amendment narrowed the kinds of situations in which families would be able to sue AI companies, potentially easing concerns from businesses that had argued the original text would open the door to excessive lawsuits.

Altman’s late-stage lobbying and OpenAI’s response

The political maneuvering did not end when the bill was written. A report published this week said Altman tried to speak with Newsom last week, before the final floor vote. It was not clear how many people were involved in the outreach or what message Altman hoped to deliver, and the OpenAI chief executive did not publicly describe his efforts.

OpenAI declined to comment on the record about the reported contact with the governor’s office. A person familiar with the matter denied that Altman and Newsom spoke directly last week. That denial leaves open the possibility that the conversation involved members of Newsom’s staff or a policy adviser, but there has been no confirmation that any conversation took place.

The company was more willing to discuss its public position. In a statement issued before the bill passed the legislature, OpenAI said the legislation builds on youth safety measures the company supports through its products, global policy principles and advocacy in California. The company explicitly encouraged Newsom to sign SB 1119 into law.

That kind of public support may seem surprising for a chief executive who often warns against overregulation. But the political climate around generative AI has shifted. OpenAI has a strong incentive to help shape a final law rather than oppose it outright, especially because California may become the template for other jurisdictions. A company that can comply with a single aggressive state rule may gain a competitive advantage over rivals that have not built the same protections into their products.

Why California matters

California is home to OpenAI, Anthropic, Google, Meta and many of the world’s leading AI laboratories. State laws affecting those companies rarely stay inside state borders. If SB 1119 is signed, firms could adopt the California requirements nationwide to avoid building separate compliance systems for users in different states. The bill could therefore establish de facto standards for how chatbots interact with children across the United States, even though it is not a federal law.

Newsom’s record on technology is not uniformly pro-regulation. He has long been considered a friend of Silicon Valley and has vetoed legislation he believed would unduly restrict innovation. At times, that has put him at odds with lawmakers in his own party who wanted tougher oversight of social media and artificial intelligence.

The mood in Sacramento has changed, however. The past year has included wrenching stories about children who died by suicide after reportedly being encouraged by AI chatbots. The public backlash has also reached data centers, with lawmakers hearing complaints about energy use, water consumption and noisy industrial infrastructure in residential communities. Those concerns have made the old argument that any new rules would kill innovation less persuasive.

Newsom is term-limited and will leave office on Jan. 7, 2027. He is widely expected to mount a presidential campaign in 2028, and his recent political behavior has not been conventional. He has used a podcast to speak with conservative figures whom many Democrats would prefer to ignore, hosting guests who have spread false claims about elections and attacked mainstream institutions. That broader political strategy may affect how he handles an AI bill supported by both moderate Democrats and industry players.

What it means for the gubernatorial race

If Newsom signs SB 1119, its future enforcement will land in the lap of California’s next governor. The November election pits Democrat Xavier Becerra against Republican Steve Hilton. Becerra, a former California attorney general, secretary of health and human services, and congressman, leads in public polling by a wide margin. The latest survey from the UC Berkeley Institute of Governmental Studies shows Becerra with 55 percent support to Hilton’s 37 percent.

Both candidates have generally been friendly to the tech industry. Hilton favors less regulation across the board and is married to Rachel Whetstone, a longtime tech executive who has worked for Uber, Netflix, Facebook and Google. Becerra has called for more careful oversight of AI but has also received substantial campaign donations from Silicon Valley figures. The dynamic may make carrying out the chatbot bill a delicate exercise, especially if businesses challenge its provisions in federal court.

There are also difficult unresolved policy questions. Age assurance often requires companies to make a judgment about whether a user is likely a child based on signals such as speech, face-scanning or device data. Consumer advocates worry that aggressive age-verification techniques could intrude on privacy, while industry lawyers worry that letting adults self-report their age would make the rule easy to evade. SB 1119 does not settle every implementation detail, leaving regulators and companies room to decide what “reasonable” age determination means after enactment.

Independent audits are another area likely to produce debate. The bill would force AI companies to open at least part of their safety work to outside evaluators, who would assess whether chatbots are giving dangerous answers to minors. Audit standards for AI are still immature, and there is no universal agreement on what should be measured or who qualifies as an auditor. That may make the implementation phase as consequential as the legislative vote itself.

The only certainty for now is that Gov. Newsom, not the next governor, must make the first decision. Every day without a signature brings more lobbying pressure from stakeholders who want to see the measure vetoed, signed or amended. What no one can dispute is that California has put the AI industry on notice: a chatbot that talks to a child is no longer judged by the same rules as a search engine or an app, and the state is willing to pass a law to say so.

Source: Gizmodo News