Amendments move Colorado AI Act's focus from risk to transparency

Colorado's landmark rules on artificial intelligence and automated decision-making technology are changing. The long-discussed reforms to the AI Act now await governor enactment after the Colorado General Assembly granted final passage to amendments that delay the law's effective date while making significant provisional changes.

Senate Bill 189 replaces the AI Act's original risk-based framework with new disclosure and transparency requirements. The shift means the law no longer carries obligations for a duty of care, risk management programs or impact assessments. It now provides for explicit disclosures around intended and harmful uses of ADMT, training data categories and deployer oversight instructions.

The law was set to take effect 30 June, a point of contention among covered entities since the law was first signed two years ago. SB 189 moves that effective date to 1 Jan. 2027.

Discussions around amendments began last year, with state lawmakers ultimately falling short on amendments during regular business and a special session called by the governor to specifically address AI Act changes. Instead, lawmakers approved the first effective date change from 1 Feb. to 30 June.

When announcing the 2025 special session, Gov. Jared Polis, D-Colo., said lawmakers needed to "work toward solutions that reduce the fiscal and negative economic impact." A spokesperson for Polis issued a public statement indicating SB 189 has achieved the governor's goals and that he will sign the bill.

"The policy framework in SB 189 is a nation-leading model created through thoughtful collaboration across many stakeholders including AI developers, those who use the technology, and consumers, moving Colorado forward," the spokesperson said. "The Governor looks forward to signing SB 189 and making Colorado a top state for innovation and entrepreneurship."

What's new?

According to Troutman Pepper Locke Partner David Stauss, CIPP/E, CIPP/US, CIPT, FIP, SB 189 "removes many of the hallmarks" from the original AI Act. He said prior references to algorithmic discrimination, non-human system disclosures and attorney general notices are among the other notable removals.

The new notice regime still presents "limited rights in narrow circumstances," Stauss told the IAPP. Those include rights to access and correct data as well as obtain a "meaningful human review" in instances of adverse ADMT decisions.

If enacted, SB 189 will come online before the Colorado's 2027 legislative session, meaning further changes will not happen before implementation. However, the bill provides for mandatory attorney general rulemaking that could bring some modifications sooner than later.

"One of the most important topics will be developing the definition of 'materially influences,'" Stauss said, "The obligations apply to ADMT that materially influences a consequential decision. So, the attorney general's rulemaking on that topic will be significant in determining the law’s eventual scope."

Changing tides

SB 189's proposed framework comes in response to stakeholder implementation concerns, but also due in part to pro-innovation sentiments sweeping the U.S.

The AI Act was originally passed while global jurisdictions sought to establish firm safety measures to curtail the potential harms stemming from AI's rapid growth. In the years since, the tone around AI shifted, allowing technologies to reach their potential and addressing safety only as necessary or where it will not stifle innovation.

"I've whittled this bill down to more of a discrimination decision bill. It's not as comprehensive and I am not happy with that," state Sen. Robert Rodriguez, D-Colo., told Colorado Public Radio. He was the original author of the AI Act and led SB 189's consideration. "But as you hear from everybody, sometimes when everybody's not happy that you're in a good place."

It is unclear how the revised framework will be received by U.S. President Donald Trump's administration, which established a Department of Justice task force to challenge state AI laws that it determined will stifle innovation. The DOJ previously joined xAI's recent lawsuit challenging the AI Act based on its algorithmic discrimination language.

The original AI Act statute provided for a legal review by a multistakeholder task force, which was delivered to state lawmakers and Gov. Polis ahead of the 2025 legislative session. The review mandate focused improving policies around anti-discrimination, bias and transparency.

However, Polis replaced the task force with his own working group after last year's legislative shortcomings. The Colorado AI Policy Work Group helped author SB 189 based on "protecting consumers, allowing innovation to thrive, and ensuring that Colorado is not left on an island."

Colorado Technology Association President and CEO Brittany Morris Saunders was a member of Polis' working group. She told the IAPP the bill before the governor represents "meaningful progress and more balanced path on AI policy."

"This bill is a compromise, shaped by intentional stakeholder engagement, and moves Colorado toward a more workable approach for startups, small businesses, employers, and technology leaders across our state," she added. "CTA will remain engaged to ensure the final law and associated rules are clear, practical, and responsive to Colorado's innovation economy."