(The Center Square) – The Colorado Senate on Friday approved guardrails around artificial intelligence in healthcare and with minors, as well as a near-total ban of AI in therapy.\u00a0<\/p>\n
Three bills, which all passed on their second reading, came amid widespread fears surrounding the rapid and largely unregulated AI boom, and in defiance of an executive order by President Donald Trump telling states to keep AI unregulated.<\/p>\n
\u201cIt is important that we\u2019re capturing and doing justice, not only for the family, but again creating very powerful policy that will be the strongest in the country,\u201d Sen. Iman Jodeh, D-Arapahoe County, said on the Senate floor about House Bill 26-1263<\/a>, the minors protection bill\u00a0she sponsored. \u201cI want Colorado to continue to set that bar for the country and make sure that kids will not be harmed.\u201d<\/p>\n The three AI bills were all approved by a voice vote, although not unanimously. They will all go to one final Senate vote before being sent to Democratic Gov. Jared Polis\u2019 desk to potentially be signed into law.<\/p>\n Colorado\u2019s legislation aimed to address different points of concern about AI use. The Psychotherapy Artificial Intelligence Restrictions bill, HB26-1195<\/a>, would ban AI models from \u201ctherapeutic communications\u201d with clients without real-time professional oversight and ban the models from generating any sort of therapeutic recommendations or treatment plans. It would allow AI agents to provide therapists supplemental or administrative help.<\/p>\n The therapy bill comes after a series of deaths by suicide<\/a> of people that had confided in AI chatbots<\/a>. In one case testified before Congress, a 16-year-old boy had told ChatGPT his suicidal thoughts. In response, the AI model told him to not tell his parents and offered to write his suicide note, according to the testimony.<\/p>\n \u201cWhat this bill comes down to is very simple. If you are in the state of Colorado receiving therapy or counseling, that that is done with a licensed professional,\u201d said HB26-1195 sponsor Sen. Kyle Mullica, D-Adams County. \u201cA chatbot or artificial intelligence isn\u2019t providing that counseling or that therapy \u2026 I think it\u2019s a commonsense bill.\u201d<\/p>\n The Conversational Artificial Intelligence Service Operator Requirements bill, HB26-1263<\/a>, would set up additional guardrails for minors who use chatbot AI models, such as ChatGPT. The bill would require AI companies to make “technically feasible measures\u201d to prevent models from creating sexually explicit content that could create emotional dependence on the model.\u00a0<\/p>\n Much like the proposed AI therapy bill, the minors bill comes shortly after a 2025 survey<\/a> by the Center for Democracy and Technology that found one-in-five high schoolers in the U.S. know somebody who had been in a romantic relationship with an AI chatbot, or had one themselves. Forty-two percent of the students said they knew somebody who had used AI for companionship or they used it themselves for friendship.<\/p>\n Colorado\u2019s minors bill would also require AI companies to make a potential suicide protocol and prohibit the use of rewards or points for accounts of minors that might encourage engagement. Violations of any measure in the bill would be punishable by a $5,000 penalty.<\/p>\n Mullica said he spoke with a neighbor whose young daughter, Juliana, died by suicide after being sexually exploited by a chatbot and read the neighbor’s emotional testimony on the Senate floor.<\/p>\n \u201cShe\u2019ll never be able to say hi to Juliana. She won\u2019t be able to walk her down the aisle, experience life with her \u2013 and I can\u2019t help but think of what that means and try to put myself into her shoes,\u201d the senator said.<\/p>\n The Use of Artificial Intelligence in healthcare bill, HB26-1139<\/a>, would ban AI used by healthcare-related companies or groups from making decisions based solely on group data. Instead, the models would have to determine healthcare decisions from medical or clinical history, the individual patient\u2019s circumstances or other factors.<\/p>\n AI models would be allowed to make expedited approvals in healthcare under the bill, but denials or delays of coverage would have to be reviewed by a professional. Any use of AI models in healthcare would have to be reported to the government.<\/p>\n The healthcare bill acknowledged the value of using AI in healthcare but said overreliance risked increasing bias that already exists in healthcare. The bill named class, ethnicity and immigration status as likely sources of bias, and added that \u201c[AI] must not replace human judgment.\u201d<\/p>\n \u201cThis bill ensures that AI can support healthcare, but it can\u2019t replace the human judgement, accountability and compassion that patients deserve,\u201d said bill sponsor Sen. Lindsey Daughtery, D-Adams and Jefferson counties, at Friday\u2019s hearing.<\/p>\n The three bills will go to their third and final Senate vote ahead of regular legislative session ending on May 13, despite an executive order<\/a> in December 2025 by Trump that threatened lawsuits at states that regulate AI. The executive order stressed that AI should be expanded and not regulated because \u201cWe remain in the earliest days of this technological revolution and are in a race with adversaries for supremacy within it.\u201d<\/p>\n Trump argued state laws would make patchwork of AI regulations and demanded the country wait until Congress creates a \u201cminimally burdensome national standard.\u201d\u00a0<\/p>\n