Connecticut's New AI Law Requires Employer Action by October
Connecticut employers must prepare for one of the nation's most detailed AI employment laws. Governor Ned Lamont will sign Senate Bill 5, the Connecticut Artificial Intelligence Responsibility and Transparency Act, creating obligations that begin October 1, 2026, with additional requirements taking effect October 1, 2027.
The law targets what it calls "automated employment-related decision technology"-a broad definition covering hiring platforms, resume screening software, assessment tools, scheduling algorithms, and performance analytics. Routine software like word processors and email are excluded.
Four Core Obligations for Employers
1. Disclose when AI drives decisions
Starting October 1, 2027, employers must inform applicants and employees in plain language when they interact with automated decision tools. The disclosure requirement becomes more detailed when the tool plays a "meaningful" role in an employment outcome.
When an automated tool generates output that serves as a substantial factor in a decision, employers must provide written notice before the decision takes effect. The notice must identify the tool, explain its purpose, list the categories and sources of personal data analyzed, describe how that data is assessed, and provide employer contact information.
2. Allocate responsibilities with vendors
The law distinguishes between "developers" (companies that build the tools) and "deployers" (employers using them). Developers must provide deployers with sufficient information to enable compliance with the law's requirements.
Developers and deployers can contract to shift notice obligations, but those allocations must be explicit in writing. Employers should review vendor contracts now to confirm who bears which responsibilities.
3. Protect trade secrets transparently
If an employer withholds information based on trade secret protection, it must still notify the affected individual what is being withheld and the legal basis for withholding it. This prevents employers from using secrecy as a blanket excuse to avoid disclosure.
4. Report AI-related layoffs in WARN notices
Effective October 1, 2026, employers filing WARN Act notices must disclose to the Connecticut Department of Labor whether layoffs relate to AI adoption or other technological changes. Connecticut becomes one of the first states requiring this disclosure.
Using AI Is Not a Legal Defense
The law amends Connecticut's Fair Employment Practices Act to explicitly state that using an automated decision tool does not defend against discrimination claims. An employer cannot escape liability for discriminatory outcomes by blaming the algorithm.
However, employers who conduct anti-bias testing before deployment strengthen their legal position. Courts and the Connecticut Commission on Human Rights and Opportunities may consider the quality, recency, scope, and results of bias testing, as well as how the employer responded to findings. This creates an incentive to test rigorously and document results.
Beyond Employment: Other Provisions
The law also requires developers to mark and make detectable all synthetic digital content-including AI-generated audio, images, text, and video-by October 1, 2027.
It establishes restrictions on AI chatbots designed to interact with children and creates whistleblower protections for employees of frontier AI developers who report potential catastrophic risks.
A separate provision creates a state regulatory sandbox allowing companies to test new AI technologies under relaxed rules.
Four Priority Steps for Compliance
1. Inventory your tools. Map every automated system touching hiring, screening, performance management, and discipline. Determine which systems' outputs constitute a "substantial factor" in covered employment decisions.
2. Review vendor contracts. Identify which vendors provide automated decision tools. Begin negotiating responsibility allocations for notice and disclosure obligations. Confirm developers can provide the information needed for deployer compliance.
3. Build a bias testing program. Work with counsel to obtain bias-testing results, methodologies, and update schedules. Document corrective actions taken in response to findings. While not a legal safe harbor, this demonstrates good faith if disputes arise.
4. Update WARN Act protocols. Evaluate whether AI adoption or technological changes contributed to planned workforce reductions. Document your reasoning for any characterization made in a WARN notice.
With the first compliance deadline less than six months away, action should begin immediately. For legal professionals advising employers, understanding these obligations is essential to protecting clients from exposure.
For more on how AI affects employment law and practices, see AI for Legal and AI for Human Resources.
Your membership also unlocks:
