US state lawmakers highlight chatbot bills, AI's impact on mental health

U.S. states are increasingly using legislative sessions to address growing artificial intelligence safety concerns in the absence of federal policymaking. AI chatbot safety measures have been a particular focus for state legislators in recent years.

Thirty-seven states introduced bills to establish chatbot rules in 2026, according to the Future of Privacy Forum, which also produced a report highlighting the common threads and trends among the proposals. Priorities across those bills included AI disclosure obligations, mental health safeguards and restrictions on underage users' chatbot interactions.

State lawmakers who were successful in passing chatbot bills this year joined a recent FPF webinar to outline how states have taken the lead on the subject and why it has been important to tackle it now despite its evolving nature.

State Sen. Kara Warme, R-Iowa, highlighted the Conversational AI Chatbot Safety Act she helped pass in July. The law, which becomes enforceable 1 July 2027, aims to prevent potentially harmful interactions with chatbots while still allowing companies to expand their AI use.

“A chat screen on an auto insurance website is very different than a conversational AI service that you can take with you now to a restaurant in New York City and have a date, you know, with your personal digital companion,” Warme said during the roundtable conversation.

She noted the legislative debate around the law focused on balancing safety and innovation, with policymakers working with technology companies "to understand how we could write this in a way that they could implement it and that they would follow it and put these protections in. Some are just for minors, and some are for all Iowans."

Lawmakers are also wrestling with means to assess responsibility and liability for chatbot harms. A majority of enacted chatbot laws lean into state attorney general enforcement for liability while Oregon and Tennessee have established a private right of action for chatbot users to seek redress.

Oregon's Senate Bill 1546 covers AI chatbot disclosures broadly and provides for USD1,000 penalties per violation. According to the bill text, the PRA is available to individuals when they suffer "an ascertainable loss of money or property or other injury in fact."

“We felt that it was really important to give some teeth to the bill where people could sue if they had someone harmed by a company that failed to live up to the standards required in the bill,” state Sen. Lisa Reynolds, D-Ore., said.

Another patchwork

The differences between state chatbot laws, regardless of how subtle, remain a point of contention.

Similar to the conversation around comprehensive state privacy law, Warme and Reynolds highlighted the range of concerns that a developing patchwork of laws could create. Industry may struggle with inconsistent compliance requirements that could ultimately lead to some form of federal preemption, which state lawmakers would prefer to avoid if U.S. Congress seeks less protective legislation than some states are already providing.

Reynolds noted she hopes the government introduces "meaningful federal legislation" to address chatbots, though she pushed back against reservations about state laws.

"I mean, the patchwork is the second-best solution," she said. "Neither the companies nor the federal government is acting. We have to, and just by the nature of how states function, there's going to be differences."

Warme highlighted how existing state legislation around a growing technology could prove beneficial to federal policymakers. Enacted chatbot laws could provide an opportunity to determine which regulatory approaches are effectively improving user safety.

"States get to be our testing grounds for lots of different things," Warme said. "We get the chance to figure out what works for our state, and to test things out and bring that feedback."

The depth of safety measures

As users increase their use of AI chatbots, lawmakers and regulators are also focused on how chatbot design features could impact consumers’ attachment and mental health.

FPF Policy Analyst for AI Policy and Legislation Justine Gluck noted that unlike other AI systems, chatbots can remember previous conversations, personalize responses and create the impression of an ongoing relationship with users.

To prevent users from creating harmful attachments, Common Sense Media Tech Policy Advisor Brenna Leasor said regulations should consider not only "what (chatbots) say, but how they're saying it."

She noted underage users may be more vulnerable to chatbots because they can have more difficulty distinguishing between real-life and immersive experiences.

AI tools are "providing this frictionless" environment for user interactions and "an easier space to engage that they don't have to kind of work through the kinks and conflict of real life of real-life relationships," Leasor said.

Regulations and model testing safeguards may help combat risks and prevent companies from “prioritizing their own engagement optimization function" over user safety, Leasor added.

Information Technology and Innovation Foundation Policy Analyst Alex Ambrose also highlighted the need for more safety protocols, but indicated personalization itself does not necessarily create a safety risk.

"Personalization inherently isn't bad," Ambrose said, explaining when personalization is combined with gamification mechanisms, it could lead to potentially addictive features and harmful engagement.

"There is kind of a middle ground here that we should be thinking about as well and not just categorizing things on one side or the other," Ambrose said.

Several enacted chatbot laws create obligations for AI chatbots to remind users they are talking to an AI system and provide resources when conversations detect users show signs of self-harm or suicidal ideation.

Sen. Reynolds said the requirement in Oregon's law for chatbot providers to offer mental health resources is a "win," noting she and her colleagues decided to "try to find a little silver lining" that they are "hopeful" parents and children will be open to utilizing.