The first half of 2026 brought significant developments regulating the use of AI in the workplace. We have previously addressed some of these developments, including California’s new AI workplace laws on termination, layoffs, and surveillance, its “No Robo Bosses Act”, Governor Newsom’s executive order directing California agencies to study AI’s impact on the labor market, the Texas Responsible Artificial Intelligence Governance Act, and President Trump’s AI executive order targeting state AI laws.
Since then, several states have issued new rules or guidance of their own relating to automated decision-making laws, new disclosure requirements, and enforcement posture, while courts have continued to test the boundaries of AI-driven hiring decisions. This article summarizes the key developments since January 1, 2026, and closes with a practical action list for employers in affected jurisdictions.
Legislative Updates
Developments in AI workforce regulation have occurred at both the federal level, through White House initiatives and congressional proposals, and the state level, in Colorado, Connecticut, Illinois, New York, and New Jersey.
Federal Law Updates
White House’s National Policy Framework
On March 20, 2026, the White House released its National Policy Framework for Artificial Intelligence, outlining a set of legislative recommendations to inform congressional consideration of a unified federal approach to AI regulation. The Framework covers seven policy areas, including child protection, free speech, innovation, AI infrastructure, intellectual property, workforce preparation, and preemption.
On workforce development, the Framework calls for Congress to help American workers benefit from AI-driven growth rather than absorb only its disruption. It recommends incorporating AI training into existing education and workforce programs, including apprenticeships, through non-regulatory means; expanding federal tracking of AI-driven workforce shifts; and strengthening land-grant institutions’ capacity for technical assistance, demonstration projects, and youth AI programs. On preemption, the Framework calls for a single federal AI standard that protects rights and innovation while avoiding a patchwork of state laws. It recommends preempting unduly burdensome state AI laws, while preserving states’ police powers (e.g., protecting children, preventing fraud, consumer protection), zoning authority over AI infrastructure, and control over their own AI use. States would be barred from regulating AI development itself, unduly burdening lawful AI-assisted activity, or penalizing developers for third parties’ unlawful use of their models.
The Framework is non-binding, and creates no obligations for employers, but previews two developments to watch. If Congress adopts its preemption approach, employers could eventually face a single national AI standard instead of a state-by-state patchwork, though states would retain authority over consumer protection and anti-discrimination. Its workforce proposals could also expand federally supported AI training and apprenticeship pathways employers can leverage.
Great American Artificial Intelligence Act
A draft of the Great American Artificial Intelligence Act of 2026 (GAAIA), the most comprehensive federal AI framework proposed to date, was made public on June 4, 2026. Compared to prior federal proposals, GAAIA takes a narrower approach to preemption, prohibiting states from passing or enforcing laws that “specifically regulate the development” of AI models for a three-year period, but leaving states free to regulate AI use in defined contexts such as employment. Because of this carve-out, state laws targeting AI in employment would likely survive even if GAAIA passes in its current form. The bill’s Workforce title is also significant for employers: it would impose additional WARN Act disclosure obligations whenever AI serves as a “substantial factor” in a covered mass layoff.
State Law Updates
Colorado: New ADMT Framework Faces an Enforcement Stay
Colorado’s ADMT Act repealed and replaced the state’s prior Artificial Intelligence Act, establishing a new framework for automated decision-making technology (ADMT) that is used to “materially influence” “consequential decisions,” including those affecting employment or employment opportunities. Employers deploying covered ADMT are required to: (1) issue a pre-use notice; (2) where an adverse outcome results from a “consequential decision,” furnish within 30 days a plain-language explanation of the decision and the ADMT’s role in it, along with instructions for seeking further information and a summary of the consumer’s rights; and (3) upon request following an adverse outcome, allow the consumer to correct the personal data relied upon and to obtain meaningful human review and reconsideration. The statute additionally bars contract terms that would indemnify an AI developer or deployer against claims arising from discriminatory ADMT use.
Enforcement rests exclusively with the Colorado Attorney General, and violations are treated as deceptive trade practices under the Colorado Consumer Protection Act; a 60-day cure period applies where the Attorney General determines the violation is curable. The Colorado Attorney General published proposed rules on August 11, 2026 to interpret and carry out the ADMT Act, covering matters such as adverse outcome disclosures and consumer rights. The public comment period closed on October 26, 2026, and a hearing that day drew testimony from industry groups, consumer advocates, and individual consumers. Stakeholders raised implementation questions concerning when ADMT “materially influences” a consequential decision, what constitutes meaningful human review, how the proposed rules interact with existing regulatory requirements, and whether businesses will have sufficient time to comply. Although the ADMT Act’s effective date is January 1, 2027, a federal district court has stayed enforcement in xAI v. Weiser (discussed below), and the stay will continue until the court decides xAI’s preliminary injunction motion, due no later than 28 days after the final rules are adopted.
Connecticut: New AI Notices and Layoff Disclosures
The Connecticut General Assembly passed SB 5 on May 1, 2026, and Governor Ned Lamont signed it into law on May 27, 2026, as Public Act 26-15. SB 5 regulates automated employment decision-making technology (AEDT), defined as technology that processes personal data and uses computation to generate output that is a “substantial factor” (meaning a factor that “meaningfully alters” the outcome of a decision) used to “make or materially influence” an employment-related decision. The law reaches decisions based on an individual’s personal data regarding hiring, promotion, discipline, training, or a material change in employment terms, including termination, but excludes non-material matters such as changes in job tasks, hours, or work assignments, and issues of workplace health and safety, scheduling, or productivity monitoring.
As of October 1, 2026, any employer conducting a reduction in force that qualifies as a mass layoff under the federal WARN Act must provide written notice to the Connecticut Department of Labor disclosing whether the employer’s use of AI or other technology was a contributing factor. This requirement applies to any mass layoff or plant closing that triggers WARN notices, not only those involving AI-driven layoff selections.
Starting October 1, 2027, employers that deploy AEDT to directly interact with job applicants or employees must disclose that fact in plain language, unless “a reasonable person would deem it obvious.” Because the phrase is undefined, employers should lean toward disclosure until further guidance is issued. Employers must also provide job applicants and employees a pre-use written notice stating:
- that the employer uses automated employment-related decision technology;
- the purposes for which the employer is using the technology and the nature of the employment-related decision that will be affected by its use;
- the trade name of the particular automated employment-related decision technology;
- the categories of personal data concerning the job applicant or employee that the technology will analyze or process, and how the personal data will be assessed in reaching a decision; and
- the employer’s contact information.
SB 5 also amends the Connecticut Fair Employment Practices Act to provide that use of an AEDT is not a defense to a discrimination complaint, though courts may consider evidence of anti-bias testing or similar proactive efforts. A violation is an unfair or deceptive trade practice under Connecticut’s Unfair Trade Practices Act, enforced solely by the Attorney General, who must provide notice and a 60-day cure opportunity for violations occurring on or before December 31, 2027.
Illinois: AI Discrimination Restrictions and Unsettled Notice Rules
Effective January 1, 2026, Illinois HB 3773 amended the Illinois Human Rights Act (IHRA) to make it a violation for an employer’s use of AI in an employment decision, including recruitment, hiring, promotion, discipline, or discharge, to have a discriminatory effect. The law broadly defines “artificial intelligence” as a “machine-based system that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations, or decisions,” including generative AI. It requires employers to notify employees and applicants when AI is used in an employment decision, and bars using zip codes as a proxy for protected class. The IDHR has enforcement authority, and individuals may pursue claims through the IHRA complaint process, with civil action available in some cases. The statute directs the IDHR to adopt implementing rules, including on when and how notice must be given.
The scope of the required notice remains unsettled: the IDHR’s May 15, 2026 proposed regulations, which would have required employers to identify the AI product and developer, the decision affected, the purpose of using AI, the data categories collected, and the job positions involved, were withdrawn in June 2026, with the related hearing canceled and no revised timeline announced. In the interim, employers should treat the withdrawn regulations as the best available compliance roadmap while monitoring the IDHR for an updated proposal.
New York: NYC Bias Audits and Continued Regulatory Scrutiny
Local Law 144 covers employers and employment agencies that rely on an automated employment decision tool (AEDT) to substantially assist or replace human judgment in screening job candidates or evaluating employees for promotion. Employers may not use an AEDT unless it has undergone a bias audit within the past year and the employer has made the audit results public, and must give candidates and employees at least 10 business days’ notice before use, explaining the qualifications and characteristics the tool evaluates and how to request an alternative selection process or accommodation, though employers are not actually obligated to grant one. Employers must also disclose, within 30 days of a written request, what data was collected, where it came from, and how long it will be kept, unless that information is already posted on the employer’s website.
On September 30, 2026, the Department of Consumer and Worker Protection (DCWP) Commissioner Sam Levine delivered remarks at a policy forum indicating that New York City intends to stay ahead of other jurisdictions in regulating AI and algorithmic decision-making affecting workers and consumers. While no particular new rules were announced, his comments point to ongoing regulatory interest that may foreshadow further enforcement and legislative activity.
New Jersey: Disparate Impact Rules and Proposed Workplace AI Safeguards
The New Jersey Division on Civil Rights (DCR) adopted regulations, effective December 15, 2025, extending disparate impact discrimination under the New Jersey Law Against Discrimination (LAD) to automated employment decision technology. The regulations cite examples of potential algorithmic discrimination, such as tools that use an employer’s existing workforce as the candidate-search baseline or that screen based on availability. As with other LAD disparate impact claims, once a complainant pleads a prima facie case, the burden shifts to the employer to show the practice is necessary to a substantial, legitimate, or nondiscriminatory interest.
Senate Bill 4075, reported favorably by the Senate Labor Committee on October 1, 2026, would regulate electronic monitoring tools (EMT) that collect worker and applicant data, automated employment decision systems (AEDS) used in hiring and firing decisions, and automated benefit and service decision systems used by public entities. Before use, an EMT or AEDS would need to pass an independent “impact assessment” confirming compliance, including an accuracy and error-rate analysis and quantified job-loss estimates, with the report kept in a public registry. The bill would bar certain surveillance practices (e.g., monitoring workers in non-work areas or compelling disclosure of personal device passwords), using AI to infringe collective bargaining or legal rights, using AI to set productivity quotas that significantly harm worker health and safety, and using AI to discriminate based on protected characteristics or union affiliation.
Employers would also need to give employees and unions 60 days’ written notice before implementing a covered system, give individuals facing adverse decisions at least 10 days’ notice and a right to contest through human review, and maintain meaningful human oversight rather than relying exclusively on system outputs. New Jersey employers should track this bill alongside their compliance with the DCR’s finalized disparate impact rules, since the two tracks impose distinct obligations that may ultimately need to be addressed together.
Noteworthy AI Cases
In addition to the legislative updates, recent lawsuits have offered insight into the liabilities potentially facing employers (and even their vendors) that use AI in making employment decisions.
Mobley v. Workday: Potential Vendor Liability for AI-Driven Employment Discrimination
Mobley v. Workday, filed in February 2023 in the Northern District of California, has set the foundation for vendor-liability for employment discrimination claims. The lawsuit alleges that Workday’s AI-powered screening tools discriminate against applicants based on race, age, and disability, violating Title VII, the ADEA, and the ADA. Its central theory, that Workday is directly liable as an “agent” of its employer-clients rather than merely a software provider, survived dismissal in July 2024, and the court certified a nationwide collective action on the age claim in May 2025. On March 6, 2026, the court held the ADEA’s disparate impact protections extend to applicants, though it dismissed certain California state law claims and a disability claim, later re-pleaded.
On June 22, 2026, the Court largely denied dismissal of the amended complaint, holding California’s FEHA may reach Workday’s tools even for out-of-state jobs, since Workday allegedly designs and operates them from California. The Court also allowed an ADA claim to proceed based on allegations that Workday’s tools treated employment gaps associated with medical leave as a proxy for disability. A separate discovery ruling on May 28, 2026 addressed access to Workday’s data. The Court found that Workday’s bias-testing data may be privileged, but the underlying performance data remains discoverable. These rulings suggest any AI tool that evaluates applicants may expose both vendor and employer to liability, and that a vendor’s California nexus alone can trigger California law.
Kistler v. Eightfold AI: Testing Consumer-Reporting Obligations for AI Hiring Tools
Kistler v. Eightfold AI, filed January 20, 2026 in the Northern District of California, advances a theory independent of algorithmic bias. Two California applicants allege Eightfold secretly compiled personal data, including social media, location, and device activity, to generate a “likelihood of success” score for employers, making it an unregistered consumer reporting agency under the FCRA and California’s Investigative Consumer Reporting Agencies Act. The theory requires only proof that Eightfold compiled “consumer reports” without the FCRA’s required disclosure, authorization, and adverse action procedures.
The court must decide whether Eightfold is a consumer reporting agency, whether its scores are FCRA-covered consumer reports, and whether a transactional exemption applies. Briefing concluded July 9, 2026, and the court heard argument August 4, 2026, with a ruling still pending. A plaintiff win would create a new compliance track, separate from discrimination theories, for any AI tool compiling third-party applicant data.
xAI v. Weiser: Challenging Colorado’s AI Law and Staying Enforcement
xAI v. Weiser began April 9, 2026, when xAI sued in the District of Colorado to enjoin the original Colorado Artificial Intelligence Act (CAIA) before it took effect, arguing its ban on AI-caused “differential treatment or impact” violated the First Amendment, Commerce Clause, Fourteenth Amendment, and Equal Protection Clause. On April 24, 2026, xAI, the DOJ (intervening for the first time in a state AI case), and Colorado’s Attorney General jointly moved to suspend proceedings pending possible legislative replacement, and the court stayed CAIA enforcement on April 27, 2026 pending a ruling on xAI’s injunction motion.
The court required xAI to file its injunction motion within 28 days of final implementing regulations or replacement legislation, and the Attorney General agreed not to enforce or investigate until 14 days after that ruling. Colorado then repealed and replaced the CAIA with SB 26-189. The stayed litigation and adoption of final rules remain key developments to monitor. The DOJ’s intervention, its first move to limit state AI regulation, could shape AI employment law nationally.
Key Takeaways for Employers
- Monitor federal developments, but don’t wait on them. The White House’s National Policy Framework and the draft GAAIA both signal a federal push toward a national AI standard, but neither currently imposes obligations, and GAAIA’s narrower preemption approach would likely leave existing state employment-focused AI laws intact. Employers should continue complying with state and local requirements rather than assuming federal preemption will arrive or displace them.
- Inventory AI tools used across the employment lifecycle. Colorado’s ADMT Act, Connecticut’s SB 5, Illinois’s HB 3773, and NYC’s Local Law 144 each turn on whether a tool substantially assists, replaces, or materially influences an employment decision. Employers should catalog hiring, promotion, discipline, and termination tools now to determine which jurisdictions’ notice, audit, and disclosure obligations apply.
- Build (or update) pre-use notice and disclosure processes. Colorado, Connecticut, and NYC all require advance notice to candidates or employees before using a covered tool, generally describing its purpose, the data it uses, and avenues for recourse. Employers operating in multiple states should design a single, layered notice process that satisfies the most stringent applicable requirement rather than maintaining separate state-by-state notices.
- Prepare for bias audits and impact assessments. NYC’s Local Law 144 requires an annual bias audit and published summary; Colorado and New Jersey’s pending SB 4075 both contemplate independent impact assessments addressing accuracy and discriminatory effect. Employers should negotiate audit rights and documentation obligations into vendor contracts now, since many of these audits depend on data only the vendor controls.
- Preserve meaningful human review. Colorado’s ADMT Act and New Jersey’s SB 4075 both require a path to human reconsideration of adverse AI-driven decisions, and the Mobley and Kistler litigation shows courts scrutinizing how much independent judgment a human reviewer actually exercises. Employers should document that human reviewers can meaningfully override or correct a tool’s output, not merely rubber-stamp it.
- Update WARN Act and layoff disclosure practices. Connecticut’s SB 5 and the draft GAAIA both add disclosure obligations when AI is a contributing factor in a covered mass layoff, and California has enacted similar requirements. Employers conducting reductions in force should build an AI-contribution assessment into their WARN Act compliance checklist now, before an October 2026-style deadline arrives in another state.
- Expect vendor liability theories to keep expanding. Mobley v. Workday shows a vendor can face direct discrimination liability as an employer’s “agent,” and a vendor’s California nexus alone may trigger California law regardless of where the applicant is located. Separately, plaintiffs in Kistler v. Eightfold AI advance a consumer-reporting theory under the FCRA based on the alleged use of third-party applicant data, rather than claims of bias. Whether the FCRA applies remains unresolved. Employers should review vendor contracts for indemnification, audit cooperation, and data-sourcing transparency.
- Track the Colorado ADMT Act’s effective date and litigation stay closely. The Act’s enforcement remains stayed pending the outcome of xAI v. Weiser. Employers with a Colorado workforce should monitor adoption of the final rules and developments in xAI v. Weiser, and prepare for compliance when the Act becomes effective and enforcement is permitted, rather than waiting for the litigation’s final resolution.
- Watch for a broader worker-protection framing beyond discrimination. NYC Commissioner Levine’s remarks, New Jersey’s SB 4075, and Connecticut’s anti-bias-testing defense provision all reflect regulators viewing workplace AI through a lens that includes surveillance, productivity quotas, and economic displacement, not just discrimination. Employers should reassess AI-driven monitoring and productivity tools with this broader framing in mind, not just their hiring and promotion tools.