US’s First State AI Bias Law Lets Job Seekers Fight Rejections

Colorado job applicants will be entitled to challenge a company’s decision not to hire them, one of the litany of employer obligations and worker rights in the state’s sweeping new law aimed at mitigating bias in artificial intelligence.

The measure (SB 205) is the most comprehensive law to date governing US employers’ use of automated decision-making tools such as resume and job candidate screening software, employment lawyers said. The details of what the law ultimately will require when it takes effect in February 2026 hinge on the guidance or rulemaking Colorado’s attorney general is expected to issue, plus potential changes by the state legislature that Gov. Jared Polis (D) has requested.

Colorado is an early mover among state, federal, and international bodies looking to regulate the growing use of AI and algorithmic decision tools, an effort largely aimed at preventing automated discrimination. Its provisions on employment discrimination will influence how lawmakers elsewhere approach legislating in this emerging area.

Much of the Colorado law tracks with existing regulatory guidance and emerging industry best practices, said Michelle Duncan, an employment lawyer with Jackson Lewis PC in Denver and former Labor Department attorney. The law’s requirements include conducting impact assessments of AI tools at least annually to monitor for unintentional bias against employees and job-seekers based on protected characteristics such as race and sex.

But it’s an unusual twist for Colorado to require that companies provide detailed notices to consumers who are negatively affected by an AI-assisted decision, plus an opportunity to formally challenge the decision with the companies and seek human review, she said.

“That is a really big lift for employers,” Duncan said. “You may have hundreds or thousands of applicants who are screened out by an AI tool. Now you’re going to have to give notice and appeal rights to every one of those applicants.”

The notice provision is well-intentioned in requiring companies to explain “that black box of how you’re being evaluated and what you’re being evaluated for,” said Matt Scherer, senior policy counsel for workers’ rights at the Center for Democracy & Technology. But the rights to challenge an employer’s decision are too vaguely written to be useful to workers, he said. The new law requires the appeal to include human review of the decision only “if technically feasible.”

“My fear is companies will find ways to essentially automate the appeal process as well,” Scherer said.

Law’s Interpretation

Only one other major US jurisdiction, New York City, has so far successfully enacted an AI-bias law, leaving the Colorado legislators in largely uncharted territory as they crafted their statute.

Consumer and labor advocates see many of the Colorado law’s requirements, including the appeal rights, as written so weakly that they’re little more than paperwork and unlikely to be enforced, Scherer said.

It’s too early to say exactly how the law’s requirements will be interpreted, Colorado Attorney General Phil Weiser said in a May 20 interview, but his office would focus its enforcement on “flagrant” violations. His office plans to seek input from a variety of sources affected by the law in deciding how to issue guidance or regulations, he said.

When signing the bill May 17, Polis also urged state lawmakers to revise the legislation before the 2026 effective date to ease the compliance burdens and avoid chilling the innovation and adoption of AI tools.

“It’s a little ambiguous at this point given the governor’s statement,” said Jenn Betts, an employment lawyer at Ogletree Deakins and co-chair of the firm’s technology practice.

For US employers, the regulatory activity on AI bias thus far includes guidance from the US Labor Department and the Equal Employment Opportunity Commission.

Other states are considering their own legislation targeting bias by AI tools, including California where the Assembly passed AB 2930 on May 21 after trimming its state budget impact. New York state lawmakers also are considering bills that address AI bias specifically in employment settings along with companies’ electronic monitoring of workers.

‘Night and Day’

The Colorado law including its notice-and-appeal provision applies not just to employment but the use of AI in education, government services, health care, housing, insurance, legal services, and lending decisions. It imposes rules on the technology developers as well as companies using the automation tools, such as giving consumers notice upfront that AI is being used and the chance to opt out in favor of human decision-making.

New York City’s first-of-its-kind law mandating bias audits for AI tools used in employment decision-making took effect in July 2023, covering tools used for hiring and promotion decisions. But Colorado’s law goes further.

“The scope and breadth of Colorado SB 205 is night and day from New York City,” said Rachel V. See, an employment lawyer at Seyfarth Shaw LLP and a former EEOC assistant general counsel for technology.

Not only does it cover many other types of decisions such as housing and credit, but it arguably could cover a broader range of employment-related decisions including performance management, scheduling, work assignments, and decisions made based on electronic monitoring of employees, See said.

The text of the law says only that it applies to decisions involving “employment or an employment opportunity,” leaving ample room for interpretation over what’s covered.

“It could also cover terminations, selection for training, compensation,” Duncan said. “We’re certainly going to be watching to see what kind of clarification we get from the AG’s office.”

Another notable difference is that in the New York City law “there is no requirement that an employer take any action in response to the audit results,” Wendy LaManque, employment lawyer at Pryor Cashman LLP, said via email. “The Colorado law requires that AI developers must disclose to the CO attorney general and any known users of the system any known or reasonably foreseeable risk of discrimination within 90 days of its discovery.”

The Colorado law also governs use of AI tools that are a “substantial factor” in assisting with employment, housing, and other major decision-making, unlike the New York City law where the tools are only subject to the bias audit and transparency mandates if they function as a dominant factor or largely replace human judgment, Scherer said.

“That’s one of the best things about the bill,” he said. “That issue was fixed.”

Enforcement Potential

The New York City law is enforced through complaints to the city’s consumer and worker protection office. The new Colorado law’s impact is similarly softened somewhat by the lack of a private right of action, said Michael Schulman, an employment lawyer at Morrison & Foerster LLP.

Weiser’s office has sole enforcement authority under the new law, and might use it sparingly at first as employers figure out how their legal obligations work around fast-developing AI technologies.

“Perhaps regulators won’t throw the book at somebody who’s doing what they can do,” Schulman said.

But a soft enforcement approach threatens to make the law ineffective, Scherer said. Some companies are likely to resist being transparent about their AI use, if they fear publicly disclosing it will increase their chances of a workplace discrimination lawsuit.

The law also includes a rebuttable presumption that employers have satisfied their required duty of care if they follow the steps laid out such as implementing a risk management policy and conducting impact assessments, according to Betts.

For employers, she said, that translates into: “There are rules, and we know the steps to be taken, and if we follow the steps it will provide some protection.”