Utah bill requires AI companies to share safety plans for children and the public
Chong Kee Siong via Getty Images
The legislation would increase company transparency but wouldn’t ‘micromanage algorithms,’ sponsor says.
This story was originally published by the Utah News Dispatch.
When presenting a bill that would require artificial intelligence companies to create and publish public safety and child protection plans, Herriman Republican Rep. Doug Fiefia told the story of Adam Raine, a 16-year-old who died by suicide with the guidance of ChatGPT.
“This was preventable. The AI company rushed out a new version of its model to keep up with its competitor that just released theirs. That rushed version, the one that released early to stay competitive, is the exact model that Adam was using,” Fiefia told a legislative committee on Tuesday. “This was a business decision that put speed and market pressure ahead of safety, and a teenager paid the price.”
His bill would enact the AI Transparency Act, which would mandate that AI developers write and post public safety plans and risk assessments for certain models. It would also protect employees who act as whistleblowers, in addition to banning developers from making misleading statements about risks and establishing civil penalties for violations.
In essence, if there are incidents that affect Utah children, AI companies have to report them to the state.
The House Economic Development and Workforce Services Committee voted unanimously to recommend Fiefia’s proposal and it will now head to the full House for consideration.
Fiefia, anticipating opposition from tech companies, also highlighted what his bill doesn’t do.
“No content mandates, no government pre-approval, no micromanaging algorithms. It doesn’t touch development, which means it doesn’t stifle innovation,” Fiefia told the committee.
Sitting next to Fiefia was actor Joseph Gordon-Levitt, who has been pushing for more AI regulations and visited the state Capitol to ask Utah lawmakers to support the legislation.
While Gordon-Levitt described himself as a “tech enthusiast,” praising the potential of AI to boost productivity and help advance science and medicine, he also pointed out potential harms.
“It’s all about how we use it, right? So the question is, what are the principles? What are the morals that are guiding the development and the design of this technology?” he said. “And I’ll tell you, from what I’ve learned, to me, there’s only one principle at play right now, It’s making money. That’s it.”
Currently, he said, there isn’t a balance between market incentives and the public good in the AI industry, “because there are virtually no laws governing it.”
“The federal government hasn’t done anything about this yet, but thank goodness the states are stepping up,” Gordon-Levitt said.
To questions on whether or not his bill would go too far, Fiefia said he already pulled out a third-party mandatory audit and the creation of safety protocol standards. Companies have already committed to advancing safety protocols, and now the state is asking them to post them for everyone to see, Fiefia said.
The more restrictive piece, he said, is the clear addition of a “child protection plan,” which other states don’t require.
“What we’re asking them to do is just tell us how you’re going to keep our kids safe. Tell us how you’re going to keep the public safe, and then if there’s an incident, you report it,” he said.
Most people participating in public comment were in support of the bill, with some parents expressing concerns about the impacts that technologies like social media have brought to kids, like anxiety and depression.
However, other groups, like Technet, a network of tech CEOs and senior executives, said the bill “takes an overly prescriptive and untested approach that goes well beyond even states like California, New York have adopted,” by imposing additional requirements and accelerating timelines, according to Andrew Wood, the organization’s executive director for Colorado and the central U.S.
“The bill improperly conflates system-level issues such as child safety obligations with highly technical frontier model safety requirements,” Wood said. “Combining these distinct issues into a single regulatory framework renders several provisions unworkable and creates regulatory uncertainty that will discourage AI development and deployment in Utah.”
Utah News Dispatch is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Utah News Dispatch maintains editorial independence. Contact Editor McKenzie Romero for questions: info@utahnewsdispatch.com.