Colorado lawmakers admit they were wrong on AI: A cautionary tale for other states

Colorado will begin a five-day special session today to repeal the left-wing artificial intelligence law Democrat Governor Jared Polis signed last year. Colorado Democrats have admitted the Colorado AI Act was a mistake and provided other states with a cautionary tale for regulating AI, but their remedy may prove worse than the disease.

In a now-infamous signing statement, Governor Polis explained all of the problems the Colorado AI Act would create for innovators, driving them from his state. Why he nonetheless signed the bill after carefully explaining everything wrong with it is baffling, as is his calling a special session to repeal it shortly afterward.

Issues with the bill include:

• Imposing liability on developers. The Colorado AI Act makes the coders who create AI systems responsible for crimes or “harms” committed by users. The same activists who want to punish firearm manufacturers for school shootings or oil companies for hurricanes want to punish coders fraud other people commit.
• No proof of intent. As Polis notes in his signing statement, law enforcement or trial lawyers would not need to prove that someone intentionally discriminated on the basis of race, sex, or another protected characteristic when using an AI system. They only need to prove that a disparate impact was felt by different people. This is an extreme legal standard that has been shown to enrich ambulance chasers at the expense of consumers.
• Patchwork of state laws. Colorado and other states passing their own extreme AI laws creates an unnavigable patchwork for developers trying to comply with conflicting state directives. This will be especially bad for smaller developers trying to compete with big tech – large companies with cash to burn can afford the compliance costs that come from these systems. New disruptors cannot.

Remedy may be worse than the disease

In the special session starting today, the legislature will consider SB 17 as a replacement, but this bill fails to address most of the problems with the Colorado AI Act. Introduced by Senate Majority Leader Robert Rodriguez, the replacement bill would retain most of liability for developers and mandate woke disclosure rules for any product. For AI processing of housing applications, for instance, the bill requires developers to provide a detailed list of all data points their algorithm might factor into a recommendation and a list of 20 personal characteristics that “most substantially influenced the output.”

It also removes the definition of “artificial intelligence” from the current law, leaving what even qualifies under this bill unclear. This is a recipe for state bureaucrats to enforce its liability and disclosure provisions arbitrarily against any developer they don’t like or who didn’t pay them off. Protections for smaller developers from the original law are watered down as well, meaning large, established tech companies can afford to comply while start-ups and disruptors will get buried in red tape and litigation.

Blue states repeat mistakes, red states learn from them

Joining Colorado in the pessimistic race to the bottom, the California legislature passed destructive AI legislation last year. Unlike Polis, Democrat Governor Gavin Newsom actually had the courage to veto the legislation as too left-wing for even him. Like the Colorado law, last session’s Senate Bill 1047 would have imputed liability for “harmful” uses of AI to developers rather than the people who actually commit crimes with the technology. Except that under the vetoed California law one need not even commit a real crime, just an imagined “harm” from the latest progressive therapist’s handbook.

A similar bill to mitigate “bias” and “harms” as defined by New York Democrats has passed the Empire State’s legislature and awaits Governor Hochul’s signature. That Americans now rely on the good judgement of Gavin Newsom and Kathy Hochul to save us from left-wing AI regulation should make us all shudder.

Utah, by contrast, has long championed permissionless innovation and moved to protect experimentation. Its model AI policy law from last year established an Office of Artificial Intelligence Policy, which is identifying regulatory barriers to AI development for removal. It will also identify where existing law covers illegal uses of AI to avoid duplicative legislation extending existing consumer protection liability to deployers, not developers. This pro-innovation regime is sure to accelerate the movement of tech entrepreneurs from Colorado across the border to the Beehive State.

Recognizing Utah’s success, Montana recently got in on the fun with its Right to Compute law. The Treasure State’s effort affirms that Montanans have a fundamental right to acquire, possess, and use technological tools, including computational resources, under the state constitution.

Both states enacted light-touch disclosure and consumer protection provisions for AI that align with existing statute.

Like Utah and Montana, other states should learn from Colorado: don’t pass anti-innovation AI bills. You will chase tech out of your state, make it harder for your residents to access cutting-edge services, and you might have to come back and fix them later like Colorado is today. States should embrace new technologies with Utah- and Montana-style sandboxes, not Colorado’s statist regimes.