Should Colorado establish national precedent in setting rules for artificial intelligence?

Colorado Senate Majority Leader Robert Rodriguez explains his bill on artificial intelligence to his chamber on May 2.

Gov. Jared Polis, a longtime supporter of disruptive industries and technologies, must decide in the coming weeks whether Colorado wants to go it alone and become the first state to regulate artificial intelligence in a comprehensive manner.

On the final day of the legislative session Wednesday, both the House and Senate cast their final votes necessary to pass Senate Bill 205, an effort from Democratic Senate Majority Leader Robert Rodriguez of Denver to put some guardrails on a quickly evolving technology. The bill requires developers of AI systems to minimize risk of algorithmic discrimination and disclose how their system was “taught”, and it mandates deployers of AI systems for high-risk decisions to monitor to ensure against bias and inform customers of the use of AI.

Rodriguez developed the bill in conjunction with Connecticut state Sen. James Maroney, and the two ran their proposals simultaneously, so much so that Rodriguez usually made changes to SB 205 whenever Maroney was making changes in his state. But on Tuesday, Connecticut Gov. Ned Lamont essentially killed his state’s bill by saying he would veto it were it to hit his desk, leaving Colorado alone in its pursuit of overseeing the sector.

Business and technology sector leaders warn the state needs to be careful in establishing first-of-their-kind regulations on the industry so that it doesn’t drive out startup companies that could establish and grow their businesses in states with fewer restrictions. Several came before the House State, Civic, Veterans & Military Affairs Committee on Saturday and said the broad definitions and mandatory appeal provisions of the bill threaten to stunt the industry if not implemented well.

How the bill regulates artificial intelligence

SB 205 does not seek to limit the growth or use of AI, but it requires both transparency and accountability of the companies that are developing systems that use machine learning, statistical analysis and data analysis to make decisions for their deployers. The bill is less a list of dos and don’ts — if signed, it will launch rulemaking led by the Colorado Attorney General’s office next year before taking effect in February 2026 — than a general framework of issues that regulators must consider.

Colorado Senate Majority Leader Robert Rodriguez speak about his AI bill.

In fact, Rodriguez scaled back SB 205 from its original reach due to concerns of overregulation by sector leaders, and it no longer applies to general-purpose AI like that which is used by ChatGPT and other consumer-facing programs or to synthetic AI that creates images. Instead, its regulations focus solely on “high-risk” programs that make decisions involving hiring and firing, medical treatment, insurance and financial decisions like loan approvals, as he fears those are the areas where algorithmic discrimination can be most costly.

“I don’t know what kinds of businesses you should have if you aren’t willing to be accountable for the systems you design,” Rodriguez told the Senate on May 2. “I love the fear and awe of this bill. Change is always hard … But this bill moves the discussion in the direction of accountability and disclosures, and that is all that this bill does.”

As explained by cosponsoring Rep. Manny Rutinel, the bill requires developers to disclose detailed information about AI systems to deployers, provide documentation assessing discrimination risks, publish a statement summarizing the risk management of systems and inform the AG’s office of any discovery of significant risk of discrimination. It requires deployers of high-risk AI to implement comprehensive risk-management policies, conduct annual reviews to prevent algorithmic discrimination, notify consumers when AI is making consequential decisions and offer appeals processes to affected consumers.

Debate about the details in Senate Bill 205

Finally, it gives the AG’s office exclusive authority to enforce the law, declaring violators to have committed a deceptive trade practice, the Commerce City Democrat added.

Colorado state Reps. Brianna Titone and Manny Rutinel discuss their AI bill with the House.

Business groups haven’t opposed the bill so much as they either have sought amendments or sought to replace its immediate implementation with the creation of a task force that can better define the guardrails that need to be put in place as AI evolves. Their concerns center on the idea that Colorado could lose its place near the top of the national innovation-economy ecosystem — a place that former dot-com entrepreneur Polis has worked to garner it — if the new rules put too many burdens on startups.

Patrick Geselbracht — senior director of tech, data and AI with Slalom Consulting — cautioned that the bill’s defining artificial intelligence as any machine-based system that infers from inputs how to generate outputs is so broad as to capture non-AI technology. He also said the guarantee of a consumer appeals process will be burdensome to many companies using AI for decision-making, and an amendment that struck the opportunity for developers to cure algorithmic discrimination will end up punishing more firms.

“The tech industry has quite an impact on our state,” said Loren Furman, president/CEO of the Colorado Chamber of Commerce. “We want to make sure … we’re not only encouraging these businesses to stay here but to act to come to Colorado.”

Could AI regulation impact economic development?

Some critics of the bill, such as Boulder Area Labor Council President Alejandra Beatty, warn that they feel it’s not doing enough to regulate what’s likely to be a very impactful technology. She was disappointed that Rodriguez cut a provision in the bill requiring any deployer to inform consumers when they are interacting with AI rather than a human, and she said the allowance for companies to conduct internal audits to look for discrimination is too loose.

 “If what’s coming out today is primarily self-regulated by the industry, we’re going to have problems,” Sheila Traister a member of the Screen Actors Guild-American Federation of Television and Radio Artists, told the House committee. “If we codify something before we fully understand it, everybody in the room understands it’s much harder to unwind it.”

Ruthie Barko — executive director of industry group TechNet for Colorado and the Central U.S. — said all of these concerns should raise questions about whether it’s best to be a pioneer in setting new regulations or to act more cautiously and create a task force. She noted that even the crucial definition of high-risk AI in SB 205 is different from what is in a European Union law on the matter.

“We want to make sure Colorado does this right,” Barko said. “We don’t want companies to have to make a decision that they either have to leave Colorado or they’re not going to choose investing in Colorado because of this.”

A case to be out front on rules for artificial intelligence

Cosponsoring Rep. Brianna Titone said that the message of caution is understandable, but that it’s also important for government to ensure that AI is being used correctly rather than as a vehicle of discrimination based upon the inputs in creating it. Colorado got ahead of the federal government in creating right-to-repair laws in several sectors, and because that’s gone successfully, the federal government is beginning to examine a national model in that field that follows Colorado’s lead, Titone added.

“It’s difficult to make a perfect policy, especially when there’s no other policy out there to compare to and AI is a moving target,” said Titone, an Arvada Democrat who serves as chair of the Joint Technology Committee. “It keeps evolving, it keeps growing, it keeps changing … And that’s why we’re relying on a lot of rulemaking in this bill, to be sure that in the process of this bill being put into law — which doesn’t go into place for some time — that that rulemaking can catch up.”

SB 205 passed the House on a 41-22 vote, with two Republicans joining most Democrats in voting for it and six Democrats — a mix of moderates and civil-rights advocates — joining most Republicans in voting against it. Meanwhile, it repassed the Senate 28-7, with Democratic Sen. Dylan Roberts of Frisco joining half the GOP caucus in opposing it.

Michael McReynolds, senior manager of government affairs for the Governor’s Office of Information Technology, was one of the people advocating for a task force to continue considering this issue, saying the pace of change in AI requires eyes to be kept upon it.

Polis has until June 7 to sign or veto bills that came out of the Legislature.