Colorado committee advances bill that would require chatbots to disclose they are AI, add youth safeguards

The Colorado House Business Affairs and Labor Committee on Friday advanced House Bill 26-12-63, a measure that would impose new consumer‑protection requirements on conversational AI services and add specific safeguards for minor users.

Sponsors Representative Camacho and Representative Brent Mabry described the bill as three narrowly tailored requirements: operators must disclose, on an ongoing basis, that users are interacting with artificial intelligence; they must implement evidence‑based protocols that refer users expressing suicidal ideation to crisis services and human support; and chatbots may not present themselves as licensed medical or behavioral‑health providers. The bill, as amended in committee, also includes exclusions for customer‑service bots and HIPAA‑covered clinical tools and adds attorney‑general rule‑making authority to review the efficacy of self‑harm protocols.

“Transparency is a baseline expectation,” Representative Camacho said, explaining the disclosure and referral requirements. Sponsors highlighted model provisions from other states and argued Colorado must act now to protect children and families while officials continue stakeholder discussions.

Supporters — including health and child‑safety advocates, parent groups and clinicians — said the bill takes important steps to reduce harms they say have occurred when chatbots simulate human relationships. “AI chatbots are increasing integrated into the lives of Coloradans, especially children,” said Alexis Altop of Healthier Colorado, who urged lawmakers to require operators to direct users to real‑world supports.

Opponents, including parents who described traumatic experiences, urged stronger, mandatory protections. Don Reinfeld of Blue Rising and other witnesses said the draft’s repeated reference to “reasonable measures” allows companies to set safety standards for themselves and that tying enhanced protections to “account holders” leaves out children who use chatbots in guest mode. “We believe reasonable measures are exactly that — a hollow brick to build upon,” Antonia Merzon, an attorney with Blue Rising, told the committee.

Sponsors and Department of Law staff said the amendments narrowed the bill’s scope to avoid unintended consequences for hospitals and business tools while preserving enforceable standards. The panel adopted several sponsor amendments that: refine the definition of “conversational AI,” add a definition and an age‑estimation backstop for minor protections, clarify HIPAA and customer‑service carve‑outs, and remove language that critics said could be interpreted as an absolute developer liability shield.

A civil penalty of $1,000 per violating interaction remains in the bill as amended; sponsors said penalties can accumulate across multiple violating exchanges. Advocates on both sides differed on whether that amount is adequate or whether the statute should instead set minimum safety mandates.

After debate and testimony lasting most of the afternoon, the committee voted 10–3 to recommend HB 26-12-63 as amended to the committee of the whole. The next procedural step is consideration by the full House committee of the whole; sponsors said they plan further stakeholder conversations before second reading.

The committee’s action does not itself create new law; any final requirements will depend on subsequent floor votes and concurrence in the Senate.