Sponsors, advocates laud AI chatbot proposal
More than 70% of teens have used artificial intelligence chatbots at least once, according to recent research.
And with that high level of use, advocates and lawmakers are pushing a plan to ensure those systems do not subject users to harm themselves or others.The Ohio Suicide Prevention Foundation, Rep. Ty Mathews, R-Findlay, and Rep. Christine Cockley, D-Columbus, said Wednesday the Legislature needs to act swiftly to keep young Ohioans safe.
Cockley said at least four Ohio children have used AI to write suicide notes. It is why she and Mathews introduced the legislation in the first place, she said.
“By encouraging developers to use a mental health framework when building and training AI, we can ensure our digital spaces continue to be places of learning, curiosity and connection, not environments where vulnerable individuals encounter harmful guidance," she said. "Protecting children in online spaces strengthens our communities offline."
Under the measure (HB 524Track), AI systems that encourage self-harm would be subject to investigations by the attorney general’s office and penalties of up to $50,000 per violation, Cockley continued. Collected penalties would go toward the 988 Suicide and Crisis Lifeline Fund.
Tony Coder, CEO of the prevention group, said outside influences like social media and problem gambling are issues for mental health.
But based on anecdotes he has heard from parents, artificial intelligence is becoming an increasingly alarming threat.
“What scares me about AI chat bots is that research has shown that when a child asks about a mental health concern and asks about what they can be doing, only 22% of the times the answer was given that was correct,” Coder said.
The proposal from Mathews and Cockley received a third hearing Tuesday in the House Technology & Innovation Committee. No opponents submitted testimony.
What could be an issue for the bill’s future is an executive order signed by President Donald Trump preventing state governments from enforcing AI laws.
Mathews said he hopes the bill would still survive if a moratorium were in place since it is not targeting AI research and development.
Mathews said he and bill supporters remain dedicated with "moving forward with this until the very end."
“And even if there is a very end, there's still, I think, a fighting chance that this is a baseline regulation that industry understands and they welcome," he said.
As the bill moves through the committee process, sponsors said they have received feedback from AI companies — some of which have already enacted protocols to prevent harmful messages from being dispersed.
Cockley said some companies have told her they have human eyes reviewing all AI chat bot messages before they are sent. That, however, can be challenging for larger models to do, she said.
Other companies are using AI technology itself to review how the chatbots interact with users, Mathews noted.