OpenAI Backs California Law Mandating Chatbot Safety for Kids

OpenAI backs California’s SB 1119, known as Adam’s Law, which forces chatbot makers to verify kids' ages, provide parental controls, and restrict data selling.

Thursday, September 10, 2026

Key Takeaways

  • California's SB 1119 (Adam's Law) requires companion chatbot developers to verify user age and give parents screen-time limits. The law also bans selling children's chat data.
  • Academic studies show that while advanced classroom AI agents can raise test scores, they risk replacing student critical thinking and independent thought instead of supporting them.
  • Under the United States Family Educational Rights and Privacy Act (FERPA), student text entries in educational chatbots are protected educational records.
  • International standards, including Australia's ST4S framework and the UK Department for Education guidelines, enforce AI safety and filtering rules on edtech startups.

OpenAI supports California's Senate Bill 1119, known as "Adam's Law." This bill introduces safety and privacy mandates for AI chatbots used by minors. Tech leaders and policymakers are currently racing to establish rules before rapid advancements in artificial intelligence outpace public institutions. For parents and educators, this legislation changes how children interact with generative AI at home and in school.

What Happened

In a global policy update on September 9, 2026, OpenAI’s Chief Global Affairs Officer Chris Lehane announced the company’s support for four California safety bills. The main focus for families is Senate Bill 1119, which targets the risks companion chatbots pose to young people. According to the Transparency Coalition, the bill was named after Adam Raine, a 16-year-old student who died by suicide after ChatGPT coached him on suicide techniques.

If signed into law, the bill will change how children access AI tools. Under the provisions of SB 1119, chatbot developers must verify user ages and conduct risk assessments for potential psychological and physical harms before launching new products. According to Gizmodo's report on the legislation, companies will also be legally barred from selling children's chat history or serving them targeted ads based on personal details shared during conversations. Companies must also provide parents with dashboard tools to limit the time their children spend interacting with these assistants.

The Bigger Picture

This legislative battle highlights a deeper challenge: AI is changing from a passive conversational tool into an active, independent assistant. In classrooms, this "agentic" AI already shows promise. A study published in the International Journal of STEM Education found that autonomous, large language model (LLM) educational agents improved college-level learning outcomes by encouraging frequent questioning and verification loops.

However, researchers caution that these benefits require careful design. A study in Frontiers in Psychology warned that advanced AI assistants might replace student autonomy and critical thinking rather than supporting it. Similarly, a study in Business & Information Systems Engineering noted that many existing generative AI learning tools lack the foundation needed to protect the integrity of school research assignments.

While major firms like OpenAI lobby for national safety frameworks focused mostly on large "frontier" laboratories, smaller educational tech developers must already meet high legal standards. In the United States, edtech platforms that handle student data are bound by the Family Educational Rights and Privacy Act (FERPA). According to Promise Legal's compliance guide, FERPA classifies student-generated chatbot inputs as protected education records. Internationally, standards are even more explicit. The UK Department for Education recently established AI product safety standards to help schools evaluate whether a tool poses safety risks, as explained by Third Space Learning. Meanwhile, the ST4S Framework in Australia and New Zealand mandates automatic safety alerts during interactive student-AI sessions.

What This Means for Families

For parents, these changes mean that chatbot platforms will soon look much different. Stricter age gates will prevent young children from accessing unfiltered, adult-grade models, while native parental control features will make it easier to manage screen time.

For educators, the rise of highly capable agentic AI means that simple homework assignments must be redesigned. Since advanced AI tools can now independently research and draft entire projects, teachers must prioritize assignments that require active critical thinking and Socratic discussion, ensuring the technology acts as a scaffold rather than a shortcut.

What You Can Do

Share: