This story by Sara Wilson appeared on Colorado Newsline on April 25, 2024
Colorado could be one of the first states to broadly regulate generative artificial intelligence, as lawmakers consider the balance between setting guardrails on a potentially-harmful technology and stifling innovation and entrepreneurship.
Senate Bill 24-205 seeks to limit discriminatory uses of AI technology by requiring businesses to self-report instances of bias to the state.
It passed through its first Senate committee hearing on a 3-2 party-line vote on Wednesday, the same day the Connecticut state Senate approved similar legislation. The idea is for states to pass a consistent, uniform law as the federal government delays action.
“This bill isn’t trying to change the world right now,” bill sponsor Senate Majority Leader Robert Rodriguez, a Denver Democrat, told committee members. “It has always been about providing a framework for accountability for biases and discrimination [and making sure] people know when they’re interacting with [AI]. The more this evolves, less and less people know they’re touching it.”
Rodriguez is a member of a multi-state AI task force that began meeting last year. He also led the bill to create the state’s consumer data privacy law in 2021, which gives people the right to ask companies to stop collecting their personal data.
The AI bill, which was essentially rewritten through a strike-below amendment in committee, would require that deployers of the technology notify consumers that they are interacting with AI, mandate risk assessments of systems and direct companies to report known risks and incidents of algorithmic discrimination to the attorney general’s office.
Those requirements are aimed to prevent discrimination that could arise when companies use AI in hiring, housing and health care decisions, among other operations. AI relies on existing data and algorithms, and that influences how they perform and opens it up to baked-in bias. Practical examples of AI bias include computer-aided diagnosis systems returning less accurate results for Black patients because it was based on data that had an overrepresentation of white patients, or a hiring algorithm favoring men because it is drawn to words more commonly used on men’s resumes.
“It has implications for almost every area of law and policy, from telecommunications to criminal law to copyright laws, civil rights, privacy, financial regulation and so much more,” Rodriguez said.
The bill is facing opposition from the business and technology community, who argue that it is too early to regulate an ever-changing technology. They compared the current state of AI to the very early days of the World Wide Web.
“The bill in its current form will do more harm than good,” said Eli Wood, the founder of Denver-based Black Flag Design. “At first glance, it seems like it’s a sensible solution to control the impacts of this technology before it negatively impacts society. But I believe it will severely curtail the ability of small organizations like ours, and ultimately negatively impact democratizing the technology for societal good.”
Wood was one of the signatories on a letter sent to Rodriguez earlier this week in opposition of the bill. The group wrote that the requirements in the bill ignore the “technological reality” of AI and would be burdensome to Colorado small businesses and startups that use the technology, potentially driving them out of the state. The opponents instead want a task force on the issue that considers the input of local business leaders, data ethics researchers and consumer advocates.
Though many people testified in opposition to the bill on Wednesday, there is consensus that some level of AI regulation is needed. The sticking point is the extent of those regulations and when they should happen.
“Legislating in an area where technology evolves daily is inherently challenging and fraught with unintended consequences. What may seem like sensible regulation today could quickly become obsolete or counterproductive tomorrow,” said Michael McReynolds, senior manager of government affairs at the Governor’s Office of Information Technology, which has registered to lobby for amendments to the bill.
The bill would set an implementation deadline for October 2025, which Rodriguez argued is ample time for a comprehensive rulemaking process to address concerns.
SB-205 is not the only bill this year to consider the impacts of AI. The Legislature gave final approval this week to a bill that would require disclosure on political campaign communications that use AI.
The bill now heads to the entire Senate for consideration. It has just under two weeks to make it through both chambers before the legislative session ends on May 8.