As federal AI policy remains unsettled, states are moving fast to fill the gap. According to the Brookings Institution, employment and transparency are among the top categories driving state-level AI legislation.

This guide breaks down the key state laws and frameworks shaping AI use in the workplace, what they require in practice, and how HR leaders can get ahead of them.

‍

Why States Are Leading on AI Employment Law

In the absence of a comprehensive federal AI framework, states, particularly those with large technology sectors and diverse workforces, have stepped in as the primary regulators. Brookings researchers note that wealthier and more “tech-active” states are passing increasingly complex legislation, with a clear focus on transparency and accountability in employment-related AI use.

The implications for HR extend across the talent lifecycle. AI tools used for resume screening, candidate assessment, performance scoring, and employee development all fall within the scope of emerging state requirements.

‍

California: The Most Comprehensive Framework

California’s suite of AI legislation is the most extensive in the country.

AB 2013: Training Data Transparency

AB 2013 requires AI developers to publish summaries of the datasets used to train their models, including whether those datasets contain personal information or third-party intellectual property. For HR technology vendors operating in California, this creates a new disclosure obligation that buyers should expect and ask about.

HR leaders evaluating AI tools should ask vendors directly: What data was this model trained on? Has that training data been documented and disclosed as required under AB 2013?

AB 489: AI Identity Disclosure

AB 489 targets misleading AI representations in health care contexts by barring AI systems from using terms that imply licensed health-profession status when that is not true. For HR technology involving AI-led conversations, this reinforces clearly identifying AI interactions as AI-generated, not human-led.

Automated Decision-Making Technology (ADMT) Regulations

California adopted regulations governing Automated Decision-Making Technology, AI systems that replace or substantially replace human judgment in consequential decisions, including employment. Under these rules, organizations must provide a pre-use notice that explains in plain language how the AI works, what it considers, and in many cases, opt-out and access rights. In practice, this means any AI tool that scores candidates, flags performance issues, or informs promotion decisions must be accompanied by explainable logic and a documented appeals path.

AB 853 and SB 53: Content Provenance and Frontier AI Standards

AB 853 requires large online platforms to detect and label AI-generated content. SB 53 establishes a Frontier AI Framework with internal safety and whistleblower requirements for large-scale AI developers. While most HR teams won’t interact with these laws directly, enterprise AI vendors operating in California will be subject to them, and vendor compliance will increasingly appear in security and procurement questionnaires.

‍

Colorado: The First Broad AI Accountability Law

Colorado’s AI Act (CAIA), set to take effect this year, is the first comprehensive state law in the US to impose broad affirmative duties on AI developers and deployers across high-risk use cases. Employment, including hiring, performance evaluation, and promotion, falls squarely within its scope.

Under CAIA, developers of high-risk AI systems must take reasonable care to prevent algorithmic discrimination. The organizations that use these tools must provide clear written disclosure to individuals when AI is used in consequential decisions. Both developers and deployers must maintain documentation of intended use, known limitations, and performance data.

The law is currently undergoing revision, with Colorado lawmakers working to refine its scope, liability framework, and compliance requirements. While specific provisions may evolve, organizations using AI in high-stakes decisions will be expected to demonstrate reasonable care, maintain robust documentation, and be prepared to justify how their systems are used.

‍

Illinois and New York: Targeted Hiring Protections

Illinois: Artificial Intelligence Video Interview Act

Illinois enacted one of the earliest state laws specifically targeting AI in hiring. The Artificial Intelligence Video Interview Act requires employers to notify candidates before using AI to analyze video interviews, explain how the AI works and what characteristics it evaluates, and obtain candidate consent. Employers are also prohibited from sharing video interview data with third parties without consent and must delete recordings within 30 days of a candidate’s request.

New York City: Local Law 144

New York City’s Local Law 144 requires employers and employment agencies using Automated Employment Decision Tools (AEDTs) in hiring or promotion decisions to conduct annual bias audits by independent third parties and publish the results public.

Employers must provide candidates and employees with advance notice (at least 10 business days) before using an AEDT to evaluate them, unless the job posting itself already includes the required disclosure. The notice must include information about what data the tool collects.

‍

What This Means for HR Teams in Practice

Across these state laws, several requirements appear consistently.

Explainability and Transparency

Any AI-generated score, assessment, or recommendation that informs a consequential decision must be explainable. Storing the reasoning behind AI outputs is increasingly a legal and operational requirement.

Opt-Out and Human Appeal Mechanisms

Individuals should have a meaningful path to contest or opt out of AI-driven decisions. For HR teams, this means building human review into workflows, not treating AI outputs as final.

Bias Auditing and Documentation

Multiple states require documentation of AI system limitations, performance metrics, and bias testing results. Organizations should maintain internal bias testing logs and be prepared to produce them.

Vendor Accountability

Several of these laws assign obligations to both employers and the AI vendors they use. HR leaders should be asking vendors the right questions: How was this model trained, and what data was used? What bias testing has been conducted? Has training data been disclosed as required by applicable law? What human oversight features are built into the platform?

‍

Responsible AI Is Required

The regulatory landscape will continue to evolve. But the principles running through every major state framework — transparency, explainability, human oversight, and bias prevention — are durable. Organizations that build these into their AI practices now will be better positioned to make defensible, equitable decisions.

‍Colleva is built with these principles as a foundation from transparent evaluation criteria to meaningful human oversight across its talent acquisition, learning and development, and employee insights solutions. To learn more, visit colleva.com.