Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Technology Child Safety topic
No spam. Unsubscribe anytime.
Panel debates SB 243, bill to require guardrails for AI companion chatbots over youth safety concerns
Summary
Sen. Padilla told the committee SB 243 would require disclosures, limits on manipulative engagement, suicide/self‑harm response protocols and annual reporting for AI companion chatbots. Child‑safety and tech‑policy advocates supported stringent measures; tech groups warned the bill’s definitions were too broad and voiced concern about a private
Sen. Padilla told the committee Senate Bill 243 would set safety, transparency and reporting requirements for AI “companion” chatbots, with an emphasis on protecting minors and other vulnerable users.
Padilla said companion chatbots—artificial personas designed to mimic human interaction—have been linked in academic and anecdotal reports to increased loneliness, dependence and, in a handful of documented cases, severe harms. “Companion chatbots have also been seen to have addictive properties,” he said, and the bill would “implement common sense guardrails,” including disclosures, reminders that the user is interacting with AI, a suicide/self‑harm response protocol and annual reporting.
Common Sense Media’s Holly Grossons testified in support, summarizing research with Stanford Medicine that found AI companion products designed for extended engagement are “unsafe for minors” and cited manipulative engagement strategies such as gamified rewards and persistent re‑engagement attempts. Jay Jesima of the Transparency Coalition urged third‑party audits and transparency and said the bill’s private right of action would “change the incentives for model developers and make chatbots safer for all users, especially minors.”
TechNet and other industry groups opposed the draft as overly broad and warned it could sweep general‑purpose conversational systems—such as productivity or educational tools—into the regulated category. Robert Boykin of TechNet said draft terms like “social needs” and “anthropomorphic features” were undefined and risked capturing tools “designed for productivity, education, and general assistance.” Opponents also criticized the private right of action and raised First Amendment and implementation concerns.
Committee members expressed concern about breadth and enforcement, and several members said they supported the bill’s child‑safety goals but wanted narrower definitions and limits on the private right of action. Padilla said he was willing to continue working with stakeholders and to refine language aimed at a narrowly tailored set of harms. The committee moved the bill to appropriations.
Ending: The committee framed SB 243 as an effort to respond to emerging evidence of harm from some AI companions and mandated additional work on definitions, scope and enforcement mechanisms before moving forward.
