Lawmakers to Decide the Future of Colorado’s AI Law
Key highlights this week:
We’re tracking 1,090 bills in all 50 states related to AI during the 2025 legislative session.
California lawmakers return to Sacramento, ready to send a raft of AI bills to the governor this month.
Illinois’ governor signed five AI-related bills into law this week.
And Colorado enters a special session to decide how best to amend its first-in-the-nation algorithmic discrimination law before it goes into effect next year, which is the subject of this week’s deep dive.
In 2024, Colorado became the first state in the nation to enact a broad algorithmic discrimination law with SB 205. The law aimed to place guardrails on the growing use of artificial intelligence by requiring developers and deployers of “high-risk” systems to use reasonable care to prevent algorithmic discrimination and to conduct impact assessments when AI systems were used in consequential decisions. But even upon signing the bill, Gov. Jared Polis (D) acknowledged that changes would have to be made before the bill took effect in February of 2026. Colorado lawmakers were unable to find a solution earlier this year and will redouble their efforts during a special session, hoping to revise the law to mitigate potential harms while also allowing AI to flourish as an industry.
Lawmakers in Colorado returned Thursday for a special session to address budget shortfalls caused by changes to the tax code from the federal budget bill. But the session also gives them an opportunity to accomplish what they were unable to finish earlier this year — amend the AI law. Legislators have proposed four competing bills so far, each offering a different path. Two have Republican sponsors and are unlikely to gain much traction. That leaves two other bills competing for how the state’s AI regulation will move going forward.
House Bill 1009: Delay and Narrow the Scope
The simplest solution, and one advocated by Republicans, would be to delay SB 205’s effective date, which is currently scheduled to take place on February 1, 2026. A bill (CO HB 1009) sponsored by Rep. Ron Weinberg (R) would delay the effective date to August 1, 2027, and significantly narrow the scope of the bill. The proposal would limit application to only AI systems that make a “consequential decision” in employment decisions or public safety, not health care, housing, or other fields specified in the original law. It also creates a small business exemption for firms with under 250 employees or $5 million annual revenues, and an exemption for local governments with less than 100,000 residents. But with Democrats controlling both chambers, this measure isn’t expected to garner much support.
Senate Bill 8: Focus on Discrimination
Some have argued that new regulatory regimes for AI are unnecessary, and that existing laws can be applied to the new technology. The “Technology-Neutral Anti-Discrimination Clarification Act” (CO SB 8) is an example of legislation from this point of view. Sponsored by Republican Senator Mark Baisley, it repeals the existing AI law and clarifies that existing anti-discrimination laws apply regardless of whether decisions are made by humans or by digital or AI systems. Sen. Baisley has submitted the bill to the conservative American Legislative Exchange Council as potential model legislation for other states to use, and is running for governor, so this could be more of a show bill than anything that has a chance of becoming law in Colorado this year.
House Bill 1008: Disclosure to Consumers
The first of two proposals with a fighting chance of getting across the finish line (CO HB 1008) would focus on transparency in AI interactions with consumers to provide or deny education, employment, lending or credit, essential government services, health care services, housing, insurance, or legal services. The bill would delete the previous law and steer clear of imposing onerous obligations on developers or deployers. Instead, it would require an AI system to clearly and conspicuously disclose to a consumer that the interaction is with AI, and not a human. The measure, which would take effect on January 1, 2027, would impose liability for violations of anti-discrimination and consumer protection laws by AI systems and would require disclosure of contact information to consumers. The measure is sponsored by Democrats in each chamber and has attracted some Republican support, while AI regulation advocates have criticized the bill for watering down regulation.
Senate Bill 4: Transparency, not Obligations
Finally, a proposal called the Colorado Artificial Intelligence Sunshine Law (CO SB 4) would remove some of the obligations under the original law, favoring transparency measures instead. Sponsored by Sen. Robert Rodriguez (D), who steered the original SB 205 into law, the measure would remove the broad "reasonable duty of care" applied to developers that drew the ire of many industry advocates, who argued it was too vague and too broad to comply with.
Instead, developers would be required to provide for deployers:
An analysis of whether and how the intended or foreseeable uses or misuses of the system could risk violating Colorado consumer protection or civil rights laws;
A description of the steps the developer has taken to mitigate those risks;
A statement describing the intended uses and foreseeable misuses; and
Any other information necessary for the deployer to comply with the law.
Deployers would no longer be required to adopt a risk management program or complete impact assessments, and would instead be required to disclose to consumers when a system influences consequential decisions. Consumers would have a right to access and correct data used in an algorithmic decision system.
Generative AI systems would also be required to disclose to consumers that an interaction is with an AI system. Developers and deployers would be jointly and severally liable for violations unless the misuse was unforeseeable and specifically disallowed. The proposal focuses more on disclosure, rather than detailed documentation and a vague duty of care.
While the bill cuts back on obligations, it increases the scope of AI regulation to apply to more parties. Rather than apply to just “high-risk” AI systems, the Sunshine Law would apply to any “algorithmic decision system” that generates a simplified output or is capable of making predictions or recommendations used to assist, inform, or replace human decision-making, with specified exceptions. It also removes some of the exemptions in the original law, such as the exemption for small businesses. The measure passed the Senate Business Affairs and Labor Committee on Thursday.
“We’re going to address some of the issues that were brought up during the process of the task force — why are we putting the burden on the deployers, and should it be more on the developers?” argued Rep. Brianna Titone (D), who helped draft the original AI law and is backing the Sunshine Law. “They are the ones that really have the ultimate responsibility for what their AI does.”
But Republican Sen. Lisa Frizell thinks the bipartisan House measure focused on disclosure represents the consensus that industry can live with. “The chief difference is (our) bill can actually be implemented by business, while the other bill is a business killer,” she countered.
The special session has become a referendum on what role, if any, state governments should play in shaping the AI era. Bills in Connecticut and Virginia fell short of passage due to concerns that regulation would scare away business. A federal effort to preempt state regulation of AI was thwarted, but state lawmakers face increased scrutiny on how much regulation they can push. This special session will reveal whether Colorado retains elements of its pioneering law and gives other states more cover, or pares back the regulation to reflect shifting political winds.
Recent Developments
Major Policy Action
Illinois: Gov. J.B. Pritzker signed five AI bills into law this week, addressing digital replicas (IL HB 3178), deepfakes (IL HB 3851), AI’s use in education (IL HB 1859, IL SB 1920), and requiring a study on how AI cameras could improve road safety (IL SB 1507).
Indiana: A legislative AI task force met last week to discuss how AI is used in state government. The chair of the task force, Sen. Liz Brown (R), does not have a particular agenda for the group, and said she wants open discussions across a wide range of topics.
Michigan: Lawmakers sent a sexual deepfake bill (MI HB 4047) to the governor for her signature, which would create a civil action for the nonconsensual dissemination of a sexual deepfake with knowledge it will create harm.
California: Lawmakers returned from a month-long summer recess this week and begin a jam-packed, four-week sprint to adjournment on Sept. 12. Appropriations committees in both chambers have lined up with hundreds of bills — including dozens of AI-related proposals — to sort out before next Friday's deadline for fiscal committees to report bills
Notable Proposals
New York: Sen. Kristen Gonzalez (D) has introduced a bill (NY S 8484) that requires informed consent from the patient for a licensed professional to use artificial intelligence to support therapy sessions that are recorded or transcribed. New York lawmakers are done for the year, but the bill can be considered in 2026.
Texas: A bill filed (TX HB 148) during the special session would prohibit the use of artificial intelligence to score constructed responses on assessment instruments administered to public school students.
