Openai Teams Up With Kids’ Safety Group On California Ballot Measure

SAN FRANCISCO — OpenAI announced a deal with a leading kids’ online safety group Friday, teaming up with a former adversary on a ballot initiative that could let California voters decide on new rules for how young people use AI chatbots, an issue that has raised scrutiny from parents and lawmakers across the country.

The compromise — first reported by POLITICO — came after the ChatGPT maker and advocacy group Common Sense Media had filed competing ballot initiatives aiming to restrict kids’ interactions with AI. The deal avoids what could have been a multimillion-dollar battle in this year’s election, and the two sides see the new proposal as offering a path to a potential national standard for regulating how youth use the technology.

“We hope to not just take this to California but well beyond,” said Chris Lehane, OpenAI’s global affairs chief and a Democratic political veteran, during a joint press conference with Common Sense Media.

The new proposal includes requirements for AI companies to determine a user's age and implement safeguards for young people using their products, as well as limiting the sale of young people’s data, among other provisions.

“Rather than confusing the voters with competing ballot initiatives on AI, we decided to work together and to enact the strongest protections in the country for kids, teens and families,” said Common Sense Media CEO Jim Steyer.

It comes amid heightened public criticism of OpenAI and other chatbot creators facing lawsuits over kids’ interactions with artificial intelligence, including over teen suicides. Working with Common Sense Media could help shield OpenAI from some of that criticism coming from lawmakers and parents.

The measure will still need to collect enough signatures to officially make it onto the November ballot — a process that can cost upwards of $10 million and which may start as soon as next month.

There’s also a chance state lawmakers could push their own legislation on chatbots, which could negotiate the measure off the ballot — an option any opponents to OpenAI’s ballot measure may prefer as a way to set their own rules and avert fighting a pricey campaign against the tech giant.

Steyer, whose group will now be taking the lead on the compromise ballot effort, said his group has already been in discussions with leaders in Sacramento, and said “we expect the Legislature will come out very soon with a comparable package.”

“We’re going to pursue every measure, from the Legislature to the ballot,” Steyer said.

Lehane said OpenAI would be standing up a ballot committee to support the effort.

Several lawmakers are already working on legislation to restrict AI chatbots, including to impose age-check requirements on chatbot platforms and temporarily ban AI-powered toys for children under 13.

The ballot proposal seeks to build on an AI chatbot law Gov. Gavin Newsom signed last year, which aims to make so-called “companion” chatbots safer for young people and requires companies to detect, remove and respond to instances of suicidal ideation by users.

The author of that law, state Sen. Steve Padilla, applauded the compromise but said he preferred to see the Legislature tackle the evolving issue. He previously slammed OpenAI for what he called an attempt to "hijack" his law to craft the language for its initial initiative.

“While this is an important milestone, there’s more work to be done and I continue to believe this issue should be tackled by the Legislature and Governor through a public process inviting all stakeholders to participate,” Padilla said in a statement Friday. “Given the rapidly evolving nature of this technology, we also shouldn’t put this law into the Constitution as proposed in the updated initiative — doing so would create an unnecessarily high-bar to revise and update that law in the future.”

Another key lawmaker looking to regulate AI, Assemblymember Rebecca Bauer-Kahan, said in a statement that she was pleased to see the two groups teaming up.

But, she said: “The legislature’s role remains unchanged; we have both the role and responsibility to protect California’s children and to represent our constituents,” adding: “We will therefore be evaluating this proposal and considering our own actions as we deem necessary to ensure the safety of our communities.”

Steyer said his group has also been in discussions with other major tech companies about the effort, without naming any. No other major AI model makers came out in support of, or opposition to, the new initiative as of Friday afternoon.

The compromise stops short of an outright ban on chatbot usage for kids: OpenAI’s initial measure focused on young people’s interactions with chatbots, while Steyer’s first proposal was more wide-ranging, also banning cell phones in schools and requiring AI literacy education in California. Those issues were dropped from the revised version.

The state attorney general would enforce the child safety requirements and impose penalties on companies found in violation, while chatbot makers would have to publish their child safety policies and submit to outside audits.

The initiative would also prohibit AI companies from targeting advertising to children, or selling their personal data without a parent’s approval. Companies would have to provide parental controls, and offer parents the option to be notified if a child expresses an intent to harm themselves.

The language would omit some uses of AI systems, like those used only for commercial purposes by businesses, video game features that resemble chatbots and products like smart speakers.

The ballot measure could still face opposition from other kids’ safety groups that have called for even tougher measures — and are likely to bristle at having a dominant AI company like OpenAI involved in crafting such rules.

Critics may also hit out at its risk assessment and audit requirements, arguing they could be more burdensome for startups and make it more difficult for new AI companies to challenge current industry heavyweights. That could draw concern from venture capital firms like Andreessen Horowitz and Y Combinator that invest in so-called “Little Tech” firms and startups.

Steyer acknowledged walking a fine line when it comes to startups, saying he’s heard from smaller firms in the past who “want to make sure that we don't craft legislation that will put them out of business.”