A Virginia legislative commission backed proposed protections for artificial intelligence chatbot users Thursday.

The Joint Commission on Technology and Science accepted recommendations organized around six policy principles to guide legislation in the next General Assembly session. The proposals include parental controls, reminders to take breaks and protocols for responding to people at risk of harming themselves or others. They would require legislation before becoming law.

Costs and responsibility for enforcement remain unresolved.

Del. David Reid, D-Loudoun, asked whether assigning new duties to several agencies would require funding and review by legislative appropriations committees.

Commission Chair Del. Cliff Hayes, D-Chesapeake, said members were voting on policy principles. Any costs would be addressed when legislation comes forward.

Executive Director Jodi Kuhn said the recommendations focus on chatbot features associated with harm while allowing users to make choices about their experience.

Companies would have to establish protocols for users showing signs of self-harm, suicidal thoughts or violence. Kuhn said chatbots should direct those users to crisis resources, and companies should make their policies publicly accessible.

Operators would also need to give users a clear way to report harmful interactions. Chatbots could not be marketed or present themselves in conversations as licensed professionals.

Other recommendations call for disclosures that chatbots are not human, reminders every two hours to take a break and making it easy to leave conversations or delete accounts.

Romantic or sexual content and excessive agreement with users would be limited by default. Kuhn said users could choose to enable additional features by creating an account.

For minors, the proposals include accounts linked to parents, the highest privacy settings by default and controls for parents or guardians.

A representative of the Virginia chapter of the American Academy of Pediatrics asked for stronger protections during public comment.

The group called for independent testing of safeguards and a prohibition on romantic or sexual interactions with minors. It also urged protections against chatbots encouraging children to talk only to them or keep conversations from trusted adults.

Hayes said the recommendations were a starting point. Lawmakers could consider additional protections when drafting legislation.

The proposed rules would cover chatbots based on what they can do, rather than how companies market them.

Kuhn said a conversation that begins with homework help can move into personal exchanges. Some customer service chatbots could also be covered if they are capable of conversations involving the identified harms.

Companies would submit risk assessments to the Virginia Department of Behavioral Health and Developmental Services. They would also report harmful incidents and provide annual reports on user complaints.

The recommendations would give the attorney general authority to investigate violations and impose civil penalties.

Kuhn said Virginia does not currently have a designated technical oversight agency for artificial intelligence. Commission staff are discussing oversight options with the administration.

The recommendations also call for seeking young people’s input on draft legislation and reviewing required reports after a resulting law has been in effect for one year.