Colorado’s Senate Bill 24-205, the first artificial intelligence (AI) regulatory law of its kind in the country, is slated to take effect February 1, 2026. The bill sets broad rules for “high-risk AI systems” that influence major life decisions in areas like jobs, housing, credit, healthcare, and education. It puts the onus on developers and users to take steps to prevent “algorithmic discrimination,” regularly assess how their systems perform, and report any problems to Colorado’s attorney general, giving that office unprecedented authority.
The only realistic path to fix this flawed bill is to push for an implementation delay during the special legislative session. Without it, the state risks saddling businesses and developers with an unworkable framework before anyone has had time to sort out its contradictions.
A deeply flawed bill
In its effort to prevent bias, SB-205 takes an interesting turn by carving out an exemption for AI specifically designed to enhance diversity or address historical injustices. In other words, while the law imposes strict rules to ensure AI doesn’t treat people differently based on legally protected characteristics, it gives a free pass to systems that do precisely that if the goal is to boost representation or correct perceived past wrongs.
This creates a glaring double standard—bias is outlawed unless it’s the “right” kind of bias—raising questions about whether the law’s commitment to neutrality is undermined by its own exceptions. It’s not clear such provisions will pass constitutional muster
On June 13, 2024, the bill’s own sponsor, Senate Majority Leader Robert Rodriguez, called for significant revisions to SB24-205 before it takes effect. He joined Governor Jared Polis and Attorney General Phil Weiser in signing a letter that calls for “refining the definition of artificial intelligence systems” and warns that overly broad definitions and burdensome disclosure requirements could lead to “unintended consequences associated with its implementation.”
Rodriguez introduced Senate Bill 25-318 earlier this year to delay SB-205’s start date until January 1, 2027, but two days before the 2025 legislative session ended, he unexpectedly moved to kill the bill.
Shortly after, on May 5, Governor Jared Polis, Denver Mayor Mike Johnston, Attorney General Phil Weiser, U.S. Senator Michael Bennet, and U.S. Representatives Joe Neguse and Brittany Pettersen signed a letter urging the Colorado legislature to delay implementation of SB-205 until 2027. “This pause will allow consumer advocates, Colorado’s business community, and other states to collaborate on a balanced, future-ready framework – one that protects privacy and fairness without stifling innovation or driving business away from our state,” they wrote.
Then on May 6, as the clock ticked down on the session, Rep. William Lindstedt offered an amendment to Senate Bill 25-322 to postpone Colorado’s AI regulations by nearly a year. He called the delay “the right thing to do” and said “cooler heads need to prevail.”
Representative Brianna Titone disagreed and, with the aid of Representative Yara Zokaie, maneuvered to call for a floor vote on the amendment. With time running out, Titone filibustered SB-322 until midnight when the session ended.
Thus the issues with SB-205 remain unresolved. One of the main points of contention that everyone from Governor Polis to Senator Bennet has with this bill is the vague definitions and the broad regulatory scope. Over 45 states have considered AI regulation, and while definitions of AI differ, SB-205 is notably lacking in clarity.
Why a delay is needed
New York’s recent RISE Act ties AI regulation to systems that partake in “human-defined objectives”. A 2024 Texas bill aimed at regulating high-risk AI systems defines those systems as technologies that “enable computer systems to perform tasks normally associated with human intelligence or perception.” Tying AI regulation to human tasks narrows the scope of both the New York and Texas bills. Colorado’s bill does not do this.
Further, various Colorado bills passed in recent sessions contain inconsistent definitions, adding fuel to the call for national regulation and consistency.
Colorado’s bill also imposes extensive documentation, disclosure, and compliance requirements on developers and deployers of so-called “high-risk” AI systems, creating massive costs that disproportionately harm small businesses and startups. This bureaucratic overhead will stifle innovation, discourage investment in Colorado, and drive tech development to states without half-cooked AI regulations.
While the original fiscal note said the taxpayer cost was $5 million, Mark Ferrandino, Director of the Office of State Planning and Budgeting, recently orally readjusted the figure to $6 million.
The various bills introduced for the special legislative session starting August 21 are similarly flawed. A two-day special session is inadequate time to develop the sophisticated proposal needed. Other than repealing this law altogether, delaying implementation until at least July 2027 is necessary. That pause would give lawmakers, businesses, and consumer advocates the time to fix SB-205’s most glaring problems: the vague and overly broad definition of AI, the compliance maze it creates for developers, and the chilling effect it will have on startups and investment in Colorado.
By aligning with clearer models like those in New York or Texas, Colorado could avoid becoming an outlier with rules so ill-defined that they invite confusion and litigation.
Just as importantly, a delay would allow time to weigh whether exemptions that permit “good” bias are even compatible with a law that claims to prevent discrimination.
Colorado’s budding AI sector needs to be allowed to blossom and flourish. Narrow the scope and reach of the law and reduce the regulatory burden placed on both developers and deployers. Colorado can both protect consumers and champion innovation.
Vanessa Rutlege is communications manager at Independence Institute, a free market think tank in Denver.
