AI Regulation in Employment: What Employers Should Know

With no comprehensive federal employment-AI law in place, states and localities are regulating how employers use AI in hiring, promotion, discipline, termination, layoffs, and other employment decisions. California, Colorado, Connecticut, Illinois, New Jersey, Texas, and New York City are among the early jurisdictions to act.

I. Automated Decision-Making in Employment

California SB 947, the “No Robo Bosses Act” (eff. July 1, 2027), prohibits sole reliance on an automated decision system (ADS) for disciplinary or termination decisions and requires human corroboration when the employer primarily relies on ADS output. If the output cannot be corroborated, or the reviewer finds it inaccurate, incomplete, or misleading, the employer may not use it.  ADS is defined to include computational processes derived from machine learning, statistical modeling, data analytics, or AI that issue scores, classifications, or recommendations used to assist or replace human decision-making. Employers that primarily rely on ADS for discipline or termination must give affected employees a written post-use notice stating the ADS reliance, the human review and corroboration, a human contact, and the right to request a description of how their data was used. Violations are enforceable by the Labor Commissioner and public prosecutors, with a $500 civil penalty per violation. Once ADS use is shown, the employer must prove it did not primarily rely on ADS or complied with the Act. Retaliation against employees who exercise these rights is prohibited.

California’s CCPA automated decision-making technology regulations (ADMT compliance beginning Jan. 1, 2027) added Article 11 to the CCPA regulations. They apply to businesses subject to the CCPA that use ADMT to make significant decisions, including hiring, allocation of work, compensation, promotion, demotion, suspension, and termination, but only where the ADMT replaces or substantially replaces human decision-making, meaning no human with authority to change the decision meaningfully reviews the ADMT’s output. Covered businesses must provide pre-use notice explaining the ADMT’s purpose, personal-information processing, human involvement, and access, opt-out, and appeal rights. Employees and consumers may opt out unless they can appeal to a human reviewer authorized to overturn the decision or ADMT is used solely to assess an employee’s ability to perform work, subject to anti-discrimination requirements. Employers may offer opt-outs for specific uses but must also provide an easy-to-use “opt-out of all” option without requiring an account or unnecessary information.

Colorado SB 26-189 (applicable to decisions made on or after Jan. 1, 2027) repealed and replaced the Colorado Artificial Intelligence Act of 2024 before it took effect. The law regulates ADMT that materially influences consequential decisions and treats employees and Colorado applicants as consumers. Deployers must provide clear, conspicuous notice before use. Adverse decisions materially influenced by ADMT require written notice within 30 days describing the decision, ADMT’s role, and how to request the data used. Individuals may request access to personal data, correction of factually inaccurate data, meaningful human review, and, when commercially reasonable, reconsideration of an adverse outcome. Developers and deployers must retain records for three years. Violations are deceptive trade practices enforceable only by the Colorado Attorney General, with no private right of action and generally a 60-day cure period before suit.

Connecticut SB 5 / Public Act 26-15 (applicable to technology deployed on or after Oct. 1, 2027) requires employers to provide written notice before using automated employment-related decision technology (AERD) as a substantial factor in decisions. The notice must identify the AERD’s purpose, trade name, personal data used and its sources, assessment method, and point of contact. The Attorney General exclusively enforces the principal AERD requirements as deceptive trade practices; there is no private right of action, and violations occurring through December 31, 2027 carry a 60-day cure period. A developer may contractually assume the deployer’s notice obligations, which is worth building into vendor agreements.

New York City Local Law 144 (eff. Jan. 1, 2023) prohibits employers from using automated employment decision tools (AEDTs) to screen candidates for hire or employees for promotion unless the tool underwent an independent bias audit within the prior year and a summary of the audit is made publicly available. Employers must provide at least 10 business days’ advance notice identifying the job qualifications and characteristics the AEDT will assess and must explain how a candidate or employee may request an alternative selection process or reasonable accommodation. The Department of Consumer and Worker Protection enforces the law, with civil penalties up to $500 for a first violation and $500 to $1,500 for subsequent violations.

Illinois’s HB 3773 (eff. Jan. 1, 2026) requires notice when AI is used in covered decisions. IDHR proposed implementing amendments on May 15, 2026, but withdrew and postponed them on June 2, 2026 to allow further coordination with other state agencies. The proposal would have required annual notice to employees, notice within 30 days of adopting or substantially changing an AI process, and notice to applicants in job postings, delivered through every applicable channel (handbooks, physical postings, intranet or website, and job postings) and identifying the AI developer and product, its use, a human contact, and the right to request a reasonable accommodation. Although withdrawn, it is the best available preview of where IDHR is headed.

II. AI Discrimination in Employment

California Civil Rights Council amendments (eff. Oct. 1, 2025) amended FEHA regulations to prohibit employers from using ADS that discriminate on a protected basis. Although a stand-alone bias audit is not required, the absence or effectiveness of anti-bias testing and similar efforts may be relevant to a claim or defense. AI evaluation of voice, physical characteristics, or schedule availability may discriminate based on race, national origin, religion, or disability, though reasonable accommodations may avoid discrimination. Vendors and other agents operating ADS on an employer’s behalf are also treated as “employers” under FEHA. ADS data and other employment records must be retained for four years.

Section 13 of Connecticut’s SB 5 (eff. Oct. 1, 2026) provides that use of AERD technology is not a defense to a discrimination complaint and permits consideration of anti-bias testing. The New Jersey Division on Civil Rights adopted N.J.A.C. 13:16 (eff. Dec. 15, 2025), which similarly provides that AI screening for characteristics such as an applicant’s schedule could discriminate based on a protected class and requires such tools to provide a mechanism to request reasonable accommodations. Illinois HB 3773 likewise prohibits AI discrimination against protected classes and the use of ZIP codes as proxies for protected classes.

Texas HB 149 (eff. Jan. 1, 2026) prohibits deploying AI with the intent to discriminate against a protected class. Disparate impact alone is insufficient to establish intent. This intent requirement is a departure from other state laws that suggest neutral criteria, such as voice or schedule, may constitute discrimination. The Attorney General has exclusive enforcement authority, there is no private right of action, and the Attorney General must give written notice and a 60-day opportunity to cure before suing.

III. AI-Related Layoffs and Other Management Regulations

California SB 951 (eff. Jan. 1, 2027) amends Cal-WARN to require that, when AI causes layoffs in whole or substantial part, the notice identify affected positions and locations, automated job functions, the technology, and must be labeled a “technology displacement.” The Employment Development Department will publish summaries, making AI-related reductions public. Similarly, § 26 of Connecticut’s SB 5 (eff. Oct. 1, 2026) requires employers serving federal WARN notice to disclose to the Connecticut Labor Department whether layoffs relate to AI or other technology.

The Great American Artificial Intelligence Act of 2026 discussion draft is a House proposal released by Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) on June 4, 2026; it has not yet been introduced as a bill. Proposed § 251 would require a federal WARN notice, when AI was a substantial factor in a mass layoff, to say so and to disclose the type and use of the AI, a good-faith estimate of the share of job losses attributable to it, and any steps taken to upskill or retrain employees beforehand. 

California has also moved on workplace surveillance. AB 1883 and AB 1331 (both eff. Jan. 1, 2027) restrict AI surveillance tools that collect employees’ neural data or recognize, infer, or predict emotional states, and prohibit monitoring employees in workplace bathrooms, with penalties of up to $500 per violation. Connecticut’s SB 5 (§ 9) separately requires employers to tell employees and applicants when they are interacting with AERD technology, unless that is obvious.

IV. How Should Employers Prepare?

1. Track AI tools used in screening, interviews, scheduling, compensation, assessments, discipline, layoffs, and workplace monitoring. Record each tool’s data, output, and weight in decision-making, and retain records for required periods: four years in California for ADS data and three years in Colorado.
2. Provide jurisdiction-specific notices stating the tool, purpose, information used, role of its output, available rights, and human contact. Create tailored disclosures for pre-use, post-use, and adverse-outcome notices.
3. Have a human in the loop who can interpret the output, evaluate relevant corroborating information, change the result, and be contacted.
4. Create an AI discrimination mitigation framework that tests impact and proxy discrimination, evaluates reasonable-accommodation pathways, documents the scope, methodology, recency, and results of anti-bias testing, and establishes a system for correcting bias and discrimination.
5. Update WARN notices and keep records that identify whether AI substantially contributed to a layoff and the affected positions, locations, and job functions.
6. Review AI vendors and contracts to identify tools tracking neural, emotional, or behavioral data or using workplace surveillance, ensure vendors provide required Colorado and Connecticut notice information, and consider shifting notice obligations to developers where Connecticut permits.

Media Contact

Public Relations Contact
Kate Lenders
Senior Marketing Manager
klenders@sgrlaw.com
312-360-6478