OpenAI Backs SB 1119 as California’s Chatbot Safety Bill Awaits Newsom’s Signature

OpenAI announced its support for California Senate Bill 1119 on August 31, 2026, adding its voice to a bill that would require AI chatbot operators to verify users’ ages, submit to independent safety audits, and build in specific protections for teenagers. The bill has already cleared both chambers of the California Legislature. Now it sits on Governor Gavin Newsom’s desk, where he has until September 30 to sign it, veto it, or let it become law without his signature.

The endorsement itself is straightforward. What makes it worth examining closely is what came before it. OpenAI’s support for SB 1119 follows the collapse, just over a week earlier, of a different regulatory strategy the company had been pursuing — one that took the form of a ballot measure rather than ordinary legislation. Understanding why that effort failed, and why OpenAI pivoted back to the bill it initially seemed to be working around, adds context that a headline alone won’t capture.

What OpenAI Announced

In a blog post published under the byline of Ann O’Leary, OpenAI’s Vice President of Global Policy, the company said it “strongly supports” SB 1119 and encouraged Newsom to sign it into law. The post followed a formal letter O’Leary sent to the governor three days earlier, on August 28, laying out the same position in more detail.

The letter frames the bill as compatible with OpenAI’s own stated approach to teen safety, noting that the company has already built age-based protections into ChatGPT. It also credits Senator Steve Padilla and Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, the bill’s authors, for their work on the issue.

This is a public position, not a technical concession or a settlement. SB 1119 is not law yet, and nothing in OpenAI’s statement changes that.

What SB 1119 Would Require

SB 1119 — formally titled “Companion chatbots: children’s safety” — sets out a series of obligations for companies that operate AI chatbots used by minors. Based on the bill’s text and legislative analyses, the requirements include:

• Determining a user’s age before applying protections
• Identifying and addressing safety risks before making a product available to young people
• Undergoing independent audits, with reporting to the California Attorney General
• Protecting minors from harmful content, including material related to self-harm and sexually exploitative content
• Giving parents tools to guide and limit their children’s use of the product
• Connecting young users with crisis-support resources when serious safety risks arise
• Limiting targeted advertising to minors and restricting the sale or sharing of their personal data
• Establishing a private right of action, allowing prosecutors and harmed children to bring civil claims

For users identified as being between 13 and 17 years old, these protections would apply automatically rather than as an opt-in setting. The bill builds on SB 243, a narrower companion-chatbot law Padilla authored in 2025 that required basic disclosures — telling users they’re talking to AI — and protocols for handling suicidal ideation or self-harm content. Sources differ on SB 243’s rollout: one report describes its core disclosure requirements as having taken effect January 1, 2026, while an industry coalition’s comment letter to lawmakers described the law as not fully in effect until July 2027. SB 1119 itself, if signed, would set its own compliance deadline of July 1, 2027, for provisions including the annual risk assessments it requires — meaning it would layer additional requirements on top of a framework that is only partly settled.

Where the Bill Stands

SB 1119 passed the full California Senate by a vote of 39–0 on May 19, 2026. It cleared the Assembly on August 31, the legislature’s final day of session before adjournment. A companion measure, AB 2023, carried similar language through the Assembly earlier in the year before being folded into SB 1119, which moved forward as the primary vehicle.

That leaves one step remaining: Governor Newsom’s decision. He has until September 30 to sign the bill, veto it, or take no action, in which case it becomes law automatically under California procedure. As of this writing, he has not publicly indicated which way he’ll go.

The Backstory: A Failed Ballot Measure First

OpenAI’s support for SB 1119 is easier to understand against the backdrop of what the company was doing in the months before it.

In December 2025, OpenAI filed its own ballot initiative in California, tied closely to the existing SB 243 framework. Senator Padilla, who authored SB 243, did not read this as a friendly gesture. In a statement from his office, he described the move as an attempt by “Big Tech” to limit “commonsense regulation” and to slow further efforts to protect children from AI chatbots — his characterization, not an independently verified account of OpenAI’s intent, but a notable marker of how the relationship between the two sides was framed at the time.

Then, in January 2026, the dynamic shifted. OpenAI and the child-advocacy group Common Sense Media announced they were merging two competing ballot efforts into a single proposal: the Parents & Kids Safe AI Act. Common Sense Media’s founder and CEO, James Steyer, welcomed OpenAI’s support publicly, describing the combined measure as the strongest youth AI safety effort in the country and framing the merger as a way to avoid confusing voters with rival proposals on the same ballot.

That combined measure needed roughly 546,651 registered-voter signatures by August 10, 2026, to qualify for the November ballot. It didn’t get them. California’s Secretary of State announced on August 21 that the petition had failed for insufficient signatures. The ballot route was dead.

OpenAI sent its letter backing SB 1119 to Newsom seven days later, on August 28, followed by a public blog post on August 31. Rather than continuing to pursue a voter initiative — a path that, if successful, would have been harder for lawmakers to amend later than an ordinary statute — the company redirected its support to the legislative bill already moving through Sacramento.

How This Compares to SB 1047

It would be a mistake to read OpenAI’s SB 1119 support as a broad reversal on AI regulation. The company’s posture looks different depending on what kind of regulation is on the table.

In 2024, OpenAI opposed SB 1047, a much broader California bill aimed at mitigating catastrophic risks from advanced AI models — the kind of systems that don’t yet exist but might pose severe harms if built carelessly. That bill would have applied primarily to large, compute-intensive models and included provisions like third-party compliance audits. Newsom vetoed it in September 2024, saying it focused too narrowly on model size rather than on where and how a system was actually deployed.

SB 1119 is a different category of bill entirely. It doesn’t address frontier-model risk in the abstract; it targets a specific, already-deployed product category — companion chatbots — and a specific population — minors. OpenAI’s support for one and opposition to the other reflects a consistent distinction in the company’s public stance: resistance to broad, model-size-based regulation, paired with willingness to back narrower, use-case-specific rules. Notably, in the same week it endorsed SB 1119, OpenAI also called for a separate bill, SB 53 — a frontier AI transparency law it had once opposed — to be strengthened further, suggesting an active recalibration of its regulatory positioning across multiple fronts at once rather than a single isolated shift.

Not Everyone in Child Safety Agrees

Coverage of OpenAI’s endorsement can make it sound as though the child-safety community speaks with one voice on SB 1119. It doesn’t.

Common Sense Media supports the bill, as does Encode and the Center for AI and Digital Policy — though the latter has asked for additional amendments, including a ban on unpredictable engagement rewards designed to hook young users and a prohibition on chatbots using design features that simulate intimacy or emotional dependency with minors. A teachers’ union, CFT, has said it would support the bill only if it were amended to explicitly bar chatbots from simulating emotional intimacy or friendship with children.

More strikingly, the Children’s Advocacy Institute at the University of San Diego School of Law is listed in opposition to SB 1119 in the California Senate Judiciary Committee’s official bill analysis. The public record doesn’t spell out the group’s specific objections in the material reviewed for this article, so it would be a mistake to assume its reasoning mirrors the industry opposition described below. It’s simply worth noting that a group with a child-advocacy focus opposed the bill — a detail that complicates any narrative of unanimous advocacy support.

On the industry side, opposition comes from trade associations rather than individual AI labs: the California Chamber of Commerce, TechNet, the Computer & Communications Industry Association, the Software & Information Industry Association, and the Civil Justice Association of California are all listed in opposition in committee records, generally on compliance-cost and regulatory-overlap grounds. No public statement from Anthropic, Google, Meta, or Mistral AI specifically addressing SB 1119 turned up in reporting on the bill; their positions, if any, are not publicly disclosed.

What OpenAI’s Product Side Looks Like Today

The policy fight connects to a product OpenAI has already built. ChatGPT for Teens is a distinct, age-gated version of the chatbot for users identified as 13 to 17 years old. According to OpenAI, the system estimates a user’s age, and if it predicts someone is under 18 — or the user states they’re between 13 and 17 — they’re automatically placed into this version rather than the standard ChatGPT experience.

The teen version includes features aimed at learning, such as Quizzes, Learning Visualizations, and a Study Mode built with input from educators. It also includes “Quiet Hours” and “Study Hours,” along with stronger default protections in sensitive areas including self-harm, violence, and sexual content. OpenAI has also published what it calls Under-18 Principles as part of its public Model Spec — rules stating that its models should not engage in romantic interactions with minors, encourage emotional dependence, or claim to be human or sentient. Parents can link accounts, adjust settings, and receive notifications in situations OpenAI considers high-risk.

These are the company’s own descriptions of its product. No independent testing or third-party audit of how well these safeguards work in practice was found in the available reporting, so their real-world effectiveness remains, for now, an open question rather than a demonstrated result. Similarly, OpenAI’s claim that nearly nine in ten teen users turn to ChatGPT weekly for learning or productivity purposes comes from the company itself and hasn’t been independently corroborated.

What Happens Next

The immediate question is simple: what Newsom does before September 30. His record on AI-related bills gives some indication of his general approach — he vetoed the broader SB 1047 in 2024 while signing narrower measures like SB 243 and SB 53 in 2025 — but that history doesn’t predict the outcome here, and nothing in the public record confirms his intentions on SB 1119 specifically.

What the record does show clearly is a company adjusting its regulatory strategy in close to real time. OpenAI spent much of 2024 and 2025 resisting broad state-level AI rules, then spent early 2026 trying to shape a narrower rule through a ballot measure of its own choosing, and only turned back to supporting the legislature’s version once that alternative path collapsed. Whether that reflects a durable shift toward engaging with targeted child-safety regulation, or a more situational response to one failed initiative, isn’t something the available evidence can settle yet. What’s verifiable now is the sequence of events — and the fact that, as things stand, SB 1119’s fate rests with one person’s decision over the next several weeks.

FAQ

What is California SB 1119?
SB 1119 is a California bill, authored by State Senator Steve Padilla with Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, that would require companion chatbot operators to verify users’ ages, undergo independent safety audits, protect minors from harmful content, offer parental controls, connect at-risk users to crisis resources, and limit advertising and data use involving children. It builds on SB 243, an earlier 2025 California law covering chatbot disclosures and self-harm protocols.

Has Governor Newsom signed SB 1119 into law?
Not as of this writing. The bill passed both chambers of the California Legislature by August 31, 2026, and Newsom has until September 30, 2026, to sign it, veto it, or allow it to become law without his signature.

What does OpenAI’s ChatGPT for Teens do?
It’s a separate, age-gated version of ChatGPT for users identified as 13 to 17. It applies stronger default protections around topics like self-harm and sexual content, includes learning-focused tools like Study Mode, and gives parents controls over settings and high-risk notifications. These are OpenAI’s own descriptions of the product; independent verification of their effectiveness wasn’t found in available reporting.

Why did OpenAI’s earlier ballot measure with Common Sense Media fail?
The proposed initiative, the Parents & Kids Safe AI Act, needed about 546,651 valid voter signatures by August 10, 2026, to qualify for the November ballot. California’s Secretary of State announced on August 21, 2026, that the petition had come up short.

How is SB 1119 different from SB 1047?
SB 1047, which OpenAI opposed and Newsom vetoed in 2024, targeted catastrophic risks from large, frontier-scale AI models generally. SB 1119 is narrower, focusing specifically on companion chatbots and child safety rather than model size or frontier-risk scenarios.

Do all child-safety groups support SB 1119?
No. Common Sense Media, Encode, and the Center for AI and Digital Policy support it, some with requested amendments. The Children’s Advocacy Institute is listed in opposition in the California Senate Judiciary Committee’s bill analysis, and a teachers’ union has said its support depends on further amendments.

When would SB 1119 take effect if signed?
Based on the bill’s text, operators would be required to comply with its provisions, including annual risk assessments, by July 1, 2027.