Artificial intelligence has steadily moved into a front-line decision-making role in the workplace. Connecticut’s newly enacted AI law reflects that and makes clear that, when technology influences employment decisions, employers must be ready to explain it.
Signed into law on May 29, 2026, Senate Bill (SB) 5 measure places new limits and disclosure obligations on employers using AI-driven tools in hiring, discipline, and termination decisions. At the core of Connecticut’s law is a broad transparency requirement. Beginning on October 1, 2027, employers must notify employees and applicants when they use “automated employment-related decision technology” (AEDT) and when that technology “materially influences an employment-related decision.” This includes tools that meaningfully shape outcomes, even where a human ultimately makes the final call. At the same time, however, the AEDT definition has significant carveouts. It excludes common tools such as word processing, spreadsheets, spam filters, and antivirus software. The AEDT definition further excludes decisions limited to scheduling, productivity monitoring, workplace health and safety, or non-material changes to an employee’s task, hours, or assignments.
When the notice must be provided depends on the interaction at issue. First, for employees and applicants who interact with the technology, an employer must inform them, in plain language, that they are interacting with automated technology. However, no disclosure is required where it would be obvious. Second, if an employer is going to use automated technology to make an employment-related decision, it must provide the affected employee or applicant with a written notice that includes:
- The employer has used the technology;
- Purpose of the technology and nature of the decision;
- The technology’s trade name;
- Categories of data the technology will process and how the data is assessed;
- Sources of personal data; and
- Contact information for the deployer.
Connecticut also breaks new ground when it comes to reductions in force. Beginning October 1, 2026, employers issuing WARN-type notices must disclose whether layoffs are connected to AI or other technological changes. Where AI is a substantial factor in selecting employees for a layoff, two obligations can stack: this disclosure to the Labor Department, which begins October 1, 2026, and the individual written notice to affected employees described above, which does not begin until October 1, 2027. A reduction in force carried out between those two dates would trigger the Labor Department disclosure but not yet the individual employee notice.
The law also emphasizes that using AI to make decisions does not shield employers from liability for biased decisions. Connecticut amended its anti-discrimination framework to clarify that employers remain responsible for outcomes, even when decisions are informed by algorithmic tools. While efforts like testing systems for bias may help demonstrate diligence, they do not provide guaranteed protection.
Unlike some employment statutes, enforcement authority for these notice requirements rests exclusively with the Connecticut attorney general, and violations may be treated as unfair or deceptive trade practices, with no private right of action. The anti-discrimination provisions are different, however: An employee who alleges that an automated tool produced a discriminatory outcome retains the private right of action available under Connecticut’s existing employment discrimination law. As a result, compliance failures could create regulatory exposure in addition to litigation risk.
From now until October 1, 2027, employers should use this lead time to evaluate their current practices and prepare for compliance. However, employers should not treat that 2027 date as the only deadline. The bar on using AI as a defense to a discrimination claim and the WARN-related disclosure for AI-driven layoffs both take effect a year earlier, on October 1, 2026. Practical steps include identifying where AI tools are used, assessing whether they materially influence decisions, reviewing vendor agreements, developing clear notices, and training HR and legal teams on how these systems operate. Overall, Connecticut’s law reinforces a growing trend: Workplace AI cannot operate as a black box. Employers who understand and can clearly explain their use of AI will be best positioned to manage risk in this evolving regulatory landscape.