Connecticut Passes Legislation Regulating the Use of AI in Employment Decisions

Connecticut has joined the growing number of states regulating the use of artificial intelligence (AI) in the workplace after passing legislation that Governor Ned Lamont is expected to sign into law. Among several other provisions, including regulations addressing AI interactions with minors, the Connecticut Artificial Intelligence Responsibility and Transparency Act (SB 5) creates new requirements for employers that use automated tools in employment-related decisions, including for recruiting, hiring, promotion, discipline, and termination.

SB 5 arrives amid a fast-developing state-by-state AI regulatory landscape. As reflected in our recent employment-focused AI alert,¹ employers must comply with numerous state and local laws requiring transparency, risk assessment, vendor oversight, and anti-discrimination controls when automated systems are used in employment decisions. Connecticut’s framework follows this trend directly addressing automated tools in recruiting and personnel decisions. 

What Tools Are Covered?

SB 5 regulates “automated employment-related decision processes (AERDP).” AERDP is broadly defined as “a computational process that generates any output, including, but not limited to, any constraint, rank, score, recommendation or classification, that (i) affects the outcome of an employment-related decision, and (ii) is not a de minimis factor that is relied upon in making, or in determining the material terms of, an employment-related decision.” Specifically, an AERDP includes technology and outputs such as:

• Scores, such as candidate-fit scores or performance-risk scores;
• Ranks, such as ranked applicant lists or internal promotion shortlists;
• Constraints, such as automated filters that exclude candidates from consideration;
• Recommendations, such as suggested candidates, interview decisions, or retention actions; and
• Classifications, such as “high potential,” “low risk,” “preferred candidate,” or similar labels.

Given the broad definition of AERDP, the law may apply to a wide range of systems, including:

• Resume-screening and applicant-tracking tools;
• Chatbots that collect applicant information or screen candidates;
• Video-interview analytics or structured interview scoring tools;
• Skills assessments, personality assessments, and cognitive assessments using automated scoring;
• Background-screening workflows that incorporate automated flags or rankings;
• Internal mobility, succession-planning, or promotion engines;
• Performance-management tools that generate automated ratings or recommendations;
• Scheduling, productivity, or workforce-management tools used for discipline or termination decisions; and
• Reduction-in-force tools or models that identify employees for selection, retention, or redeployment.

Key Compliance Requirements for Employers

A central feature of SB 5 is transparency. Similar to requirements in California and Illinois,² employers using an AERDP to generate any output for the purpose of making or as a substantial factor in making an employment-related decision will need to provide disclosures prior to using the tool. These notices should explain in plain language that an automated tool may be used, the nature and purpose of the tool, the personal data it processes, the outputs it generates, how those outputs may be used, and whether additional information or human review may be requested. In the event that an AERDP is used for an adverse employment-related decision, employers must also disclose to employees or applicants a “high-level statement disclosing the principal reason or reasons for such adverse employment-related decision, including, but not limited to, (A) the degree to which, and manner in which, the output generated by such automated employment-related decision process contributed to such adverse employment-related decision, (B) the type of data that were processed by such automated employment-related decision process in generating such output, and (C) the source of the data.”

SB 5 also makes anti-bias review a practical necessity. Specifically, SB 5 amends the Connecticut Fair Employment Practices Act to cover the use of an AERDP that has a discriminatory effect. The law directs the relevant state body or court to “consider any evidence, or lack of evidence, of anti-bias testing or similar proactive efforts to avoid such discriminatory practice, including, but not limited to, the quality, efficacy, recency and scope of such testing or efforts, the results of such testing or efforts and the response thereto.” SB 5’s “AI is not a defense” principle reinforces that employers remain responsible for discriminatory outcomes even when a vendor or automated tool contributed to the decision.

Employers using an AERDP should evaluate whether automated tools produce statistically significant disparities on the basis of protected characteristics. Where disparities are identified, employers should assess whether the decision is job-related and consistent with business necessity, whether less discriminatory alternatives are available, and whether mitigation measures are required.

Employers also should ensure that automated outputs do not become unreviewed final decisions. Human oversight is most effective when decision-makers understand the tool’s purpose and limitations and retain authority to question or override automated outputs, all of which should be documented.

Enforcement and Implementation Timeline

The employment-related provisions will be enforceable exclusively by the Connecticut attorney general under the Connecticut Unfair Trade Practices Act. SB 5 includes staggered effective dates beginning 1 October 2026, and because different provisions may become operative at different times, employers should develop a phased compliance plan rather than waiting for all obligations to take effect. For example, employers should: 

• Now–Q3 2026: Inventory automated tools, identify covered use cases, review vendor agreements, and assign internal owners.                       
• Q3–Q4 2026: Draft plain-language disclosures and pre-decision notices; create internal escalation and human-review workflows; begin bias testing for high-impact tools.
• Q4 2026–2027: Implement training for human resources and managers; update applicant and employee-facing materials; refresh procurement standards; monitor regulatory guidance and enforcement developments.
• Ongoing: Repeat audits, update notices after tool changes, document mitigation, and reassess tools when business practices or job criteria change.

Employers should also monitor further guidance from Connecticut regulators and the attorney general, as well as any amendments, interpretive guidance, or enforcement announcements affecting the law’s workplace provisions.

Our Labor, Employment, and Workplace Safety practice lawyers regularly counsel clients on a wide variety of topics related to emerging issues in labor, employment, and workplace safety law, and they are well-positioned to provide guidance and assistance to clients on AI developments.