States Take Different Approaches to Regulate AI in Employment

Since the beginning of 2025, state lawmakers across the country have considered numerous proposals to regulate the use of artificial intelligence in the workplace, while some legislatures have enacted broader AI laws with workplace implications. [See Artificial Intelligence Legislation Tracker (2026), Artificial Intelligence Legislation Tracker (2025) (Archived), and Artificial Intelligence State Law Survey.]

Existing federal employment discrimination laws apply when employers use AI, but in the absence of a comprehensive federal law specifically governing that use, the state proposals and enactments have sought to address discrimination and disclosure in different ways.

In an interview with the State Net Capitol Journal™, William Terrell, a Practical Guidance content manager in labor and employment law for LexisNexis®, discussed several emerging regulatory pathways reflected in proposed and enacted AI laws.

Terrell said the influx of legislative proposals and enactments should motivate compliance professionals to take a hard look at their policies now, because he expects regulation of AI in employment to become the norm throughout the nation in the not-so-distant future.

“This is a moment for compliance professionals to develop comprehensive approaches that cover a wide variety of workplace issues that AI could inform, including hiring, workplace monitoring, employee management, and adverse employment decisions,” he said.

Automated Decision-Making Bills Considered in Over Third of States in 2026

Lawmakers in at least 19 states have considered legislation this year dealing with automated decision-making, according to the LexisNexis® State Net® legislative tracking system. Four of those states have enacted such measures.

Upstream Rules Target Developers, Downstream Rules Target Deployers

Terrell said state legislators have addressed AI in employment through both upstream and downstream regulation.

“You have ‘frontier’ model or ‘advanced’ model safety that regulates the upstream deployment of AI tools by their creators and then you have the downstream regulation of how employers use those tools to make employment decisions,” he said.

He cited California SB 53 (2025) [codified as Cal. Bus. & Prof. Code § 22757.10 et seq.] and New York SB 6953 (2025) [codified as N.Y. Gen. Bus. Law § 1420 et seq.] as examples of an upstream regulatory model. Although neither law directly regulates employment decisions, both impose requirements on certain frontier-model developers whose models could be incorporated into employment technology.

Colorado lawmakers took a hybrid approach when they enacted SB 189 (codified as Colo. Rev. Stat. § 6-1-1701) this past May. The law imposes disclosure and transparency obligations on both developers and deployers of automated decision-making technology used to make consequential decisions involving employment, as well as education, housing, lending, insurance, healthcare, and certain government services.

This marks a retreat from the state’s 2024 AI Act (SB 205), which required developers and deployers of high-risk AI systems to exercise a duty of care against algorithmic discrimination in those same decisions, backed by mandatory risk assessments. SB 189 repealed that duty-of-care framework before it ever took effect, replacing it with consumer notice rights, a right to human review of adverse decisions, and fault-based apportionment of discrimination liability. (For more information about this law, see Colorado Automated Decision-Making Technology (ADMT) Act: Developer and Deployer Compliance.)

Different State Approaches to Algorithmic Bias

Terrell said AI-related employment discrimination proposals and enactments generally fall into two major categories: applications or extensions of existing civil rights laws to automated tools and AI-specific employment discrimination prohibitions.

Illinois HB 3773 (2024) [amending 775 Ill. Comp. Stat. 5/2-101 et seq.], which took effect on Jan. 1, 2026, straddles both categories. It amended the Illinois Human Rights Act to address AI in a broad range of employment actions, including hiring, promotion and discipline. It also prohibits employers from using ZIP codes as proxies for protected classes and requires notice when employers use AI for covered employment purposes.

New York City began to regulate AI in employment before its home state. Local Law 144 of 2021 (codified as N.Y.C. Admin. Code § 20-870 et seq.) prohibits employers and employment agencies from using an automated employment decision tool (AEDT) to substantially assist or replace discretionary employment decisions unless the tool undergoes an independent bias audit within the preceding year, the employer posts a summary of the audit results publicly, and candidates and employees receive advance notice that an employer will use an AEDT. (See Automated Employment Decision Tools: Frequently Asked Questions.) Other cities may follow New York City’s lead and begin regulating AI in the workplace before their home states act.

Terrell is also particularly interested in the implications of Texas HB 149 (2025) [codified as Tex. Bus. & Com. Code § 552.056], a broad AI governance law that focuses, in part, on discriminatory intent in the development and deployment of AI systems. The law prohibits a person from developing or deploying an AI system with the intent to engage in unlawful discrimination against a protected class. It also expressly provides that disparate impact alone is insufficient to establish discriminatory intent.

“Trying to grapple with intent is really interesting,” Terrell said. “Because when AI discriminates, it does so most often with implicit bias. Implicit bias is something that lawmakers don’t always feel comfortable reckoning with or thinking about.”

He added: “It’s actually refreshing to see these states think about the nature of bias. So, thinking about how a ZIP code could somehow reflect an implicit view about race or about some other protected characteristic, I think is really instructive for how these laws can be used to protect against discrimination.”

New Jersey, in contrast to Texas, draws a direct connection between AI technologies and discrimination. In December 2025, the New Jersey Division on Civil Rights issued rules specifically listing automated decision-making technology as a potential source of unlawful disparate-impact discrimination. (See N.J.A.C. 13:16.) As an example, it cited a tool trained to evaluate applicants by reference to an employer’s existing workforce that happens to be predominantly white and male. The tool might systematically favor candidates who resemble the current pool of employees, producing a disproportionately negative effect on a protected class.

To Terrell, this example is especially illustrative of one path that these laws could take. He noted that employers in New Jersey are liable for disparate-impact discrimination even if they did not develop the automated decision-making tool themselves and can only defeat that liability by showing that the practice is a business necessity. (For more information on state artificial intelligence laws, see Artificial Intelligence State Law Survey.)

More States Likely to Tackle AI in Employment in Future

Terrell predicted that AI regulation in employment would increasingly focus on transparency.

“As we learn more about how AI can rear its head in discrimination,” he said, “I think we are going to see a lot more employers having to disclose their use of AI with respect to all stages or all levels of employment decisions.”

Terrell said he expects to see more states adopt new AI laws in the labor and employment arena. He compared the emerging AI rules to other employment mandates, such as salary-transparency and paid-leave requirements, that expanded as additional states adopted their own versions.

Terrell said he’d counsel compliance personnel to be proactive rather than reactive in addressing the implications of using AI in employment, both to reduce enforcement and litigation risk under differing state and federal regulatory frameworks and to foster a sense of goodwill with employees.

“Employees, especially U.S. employees, are really interested in how AI is going to affect their jobs,” he said. “So, it’s better to be on the early end of adopting best practices concerning AI in the workplace.” (See Artificial Intelligence in the Workplace: Best Practices and Artificial Intelligence: Labor, Employment & Benefits Resource Kit.)

—By SNCJ Correspondent BRIAN JOSEPH

Visit our webpage to connect with a LexisNexis® State Net® representative and learn how the State Net legislative and regulatory tracking service can help you identify, track, analyze and report on relevant legislative and regulatory developments.