Somewhere today, Copilot is drafting an email that Copilot will later summarize – efficient, if a little uneventful. For ordinary office correspondence, that may be the point. For decisions affecting someone’s employment, however, several states are less willing to leave the human role quite so undefined.
Colorado’s first artificial intelligence (AI) law (SB 24-205) emphasized risk-management programs addressing algorithmic discrimination. Delayed and replaced before taking effect, it was followed by laws focusing more directly on how employers use AI, what they must disclose and how employment decisions can be reviewed.
Limits on Automated Discipline and Termination
California’s No Robo Bosses Act (SB 947), effective July 1, 2027, restricts reliance on automated systems in discipline and termination decisions. It prohibits sole reliance and, where an employer primarily relies on an automated system, requires a person to corroborate its output against other evidence, such as personnel records or performance data. The employer must also provide the affected employee written notice after the decision. Although the meaning of “primarily relies” remains undefined, the law makes clear that discipline and termination decisions require meaningful human involvement when automated systems are used in the decision-making process.
New Rights to Explanation and Human Review
Colorado’s SB 26-189, effective Jan. 1, 2027, addresses what happens after certain adverse AI-assisted employment decisions, such as rejecting an applicant or offering lower compensation. The law requires a plain-language explanation, within 30 days, of the technology’s influence. Affected individuals may also seek correction of inaccurate personal data and meaningful human review and reconsideration when they are commercially reasonable.
The scope of Colorado’s requirements remains unsettled. At an Oct. 26, 2026 hearing on the attorney general’s proposed implementing rules, business groups argued that the proposed definition of “material influence” could trigger notice and human-review requirements even when AI contributes to an outcome ultimately reached by a person. The Colorado Chamber of Commerce also warned that requiring notice and review for large numbers of unsuccessful job applicants could create significant burdens, particularly in high-volume hiring. Consumer advocates, meanwhile, argued that the proposed rules could unduly limit when human review is available. Separately, in X.AI LLC v. Weiser, the attorney general agreed to defer enforcement until 14 days after a ruling on a preliminary-injunction motion following final rules. These developments warrant monitoring as rulemaking and litigation proceed.
Regulations under the California Consumer Privacy Act (CCPA), beginning Jan. 1, 2027, provide related rights for certain significant employment decisions, including rights to opt out of qualifying automated decision-making and to obtain information about how the technology was used in the decision; for some decisions, a business may offer a human appeal in place of the opt-out.
Human Involvement Does Not Eliminate Discrimination Risk
The same states imposing new requirements on AI-assisted employment decisions also continue to regulate discriminatory employment practices under existing law. Since Oct. 1, 2026, Connecticut has expressly provided that using automated employment technology is not a defense to a discrimination complaint. California’s Fair Employment and Housing Act regulations address discriminatory automated decision-making, including the relevance of anti-bias testing and when vendors may act as an employer’s agent, while Illinois prohibits employers from using AI in ways that discriminate against protected classes.
Federal law provides another layer of protection. Although the Equal Employment Opportunity Commission’s June 4, 2026 enforcement plan disavows disparate-impact litigation, Title VII’s disparate-impact provision remains intact and private claims continue. In Mobley v. Workday, Inc., for example, an applicant alleges discrimination involving an AI screening system, and the case against the screening vendor remains pending.
New Disclosure Requirements for AI Use and Workforce Reductions
States are also requiring employers to disclose AI’s role in employment decisions. As we discussed in June, Illinois requires employers to notify affected employees and applicants of covered AI use, although proposed regulations specifying the notice requirements were later withdrawn. The CCPA regulations similarly require pre-use notice for covered automated decision-making, while Colorado’s SB 26-189 requires notice when individuals interact with covered automated decision-making technology. Beginning Oct. 1, 2027, Connecticut will require written notice identifying the AI tool, its purpose, the personal data used and a contact for inquiries.
Disclosure requirements also extend to automation-related layoffs. Connecticut requires employers filing notices under the federal Worker Adjustment and Retraining Notification (WARN) Act with its labor department to disclose whether the layoffs are related to AI or another technological change. Starting Jan. 1, 2027, California’s SB 951 will require qualifying Cal/WARN notices for workforce reductions caused in whole or in substantial part by AI or other automated technology to identify the affected occupations, automated job functions and technology involved.
The Bottom Line
AI can help employers evaluate information and recommend outcomes. But these state laws increasingly ask whether consequential decisions can be explained, reviewed and supported by evidence beyond an automated output. Employers should assess where AI influences employment decisions, who reviews those outputs, and what notices or records may be required. As AI becomes a more ordinary part of work, the human role in decisions about people may matter more, not less.
Please contact the authors of this blog or your BakerHostetler attorney contact if you have any questions.
