The short answer: AI in US hiring is legal, but you are liable for its outcomes. Federal law sets the floor. A growing patchwork of state and city laws stacks disclosure, audit, and impact-assessment duties on top. Here is what applies, where.
The state-by-state map (at a glance)
Federal foundation: EEOC, Title VII, ADEA, ADA
Every state and city rule sits on top of federal anti-discrimination law. There is no AI carve-out.
Title VII of the Civil Rights Act prohibits employment decisions that cause a disparate impact on race, color, religion, sex or national origin. The EEOC's 2023 technical assistance document, Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence, confirms what many hoped was ambiguous: if your AI screening tool selects protected-class candidates at less than four-fifths the rate of the top-selected group, you have a prima facie disparate impact — and you (not the vendor) are liable.
The Age Discrimination in Employment Act (ADEA) extends the same logic to candidates 40 and over. The EEOC's 2024 iTutorGroup settlement (AI rejecting female applicants over 55 and male applicants over 60) made clear that algorithmic age screening is enforceable, even when no human reviewed the decision.
The Americans with Disabilities Act (ADA) adds a second front. The EEOC's 2022 guidance warns that AI tools measuring reaction speed, personality, or video performance can screen out candidates with disabilities who could perform the job with accommodation. Staffing firms must offer alternative assessment formats on request.
NYC Local Law 144 (AEDT)
In force since July 2023, Local Law 144 was the first US city law to regulate automated employment decision tools. It applies when an AEDT is used to substantially assist or replace discretionary hiring or promotion decisions for a role performed in New York City.
- Bias audit: Independent audit within the prior 12 months, measuring selection rates by sex and race/ethnicity categories.
- Public summary: Audit results published on the employer's or agency's website.
- Candidate notice: At least 10 business days before the tool is used, with the opportunity to request an alternative process.
- Scope: Both employers and employment agencies — so staffing firms placing into NYC are covered even if the client runs the tool.
Penalties start at $500 for a first violation and reach $1,500 per day of ongoing non-compliance.
Illinois AI Video Interview Act (820 ILCS 42)
The oldest US AI hiring statute (2020) and still one of the strictest. Before using AI to analyze a video interview of an Illinois applicant, you must:
- Notify the candidate in writing that AI will be used.
- Explain, in plain language, how the AI works and the general characteristics it evaluates.
- Obtain express consent before the interview.
- Share the video only with those whose expertise is needed to evaluate it.
- Destroy all copies — including backups — within 30 days of a candidate request.
- If AI is the sole screening mechanism, report annual race and ethnicity data to the Illinois Department of Commerce.
A 2024 amendment (HB 3773) extended AI anti-discrimination duties to all employment decisions, not just video interviews, effective January 1, 2026.
Colorado AI Act (SB24-205)
Effective February 1, 2026, Colorado's AI Act is the first comprehensive US state law on high-risk AI — and hiring is explicitly a high-risk use case. Staffing firms acting as deployers must:
- Complete an impact assessment for each high-risk hiring system, updated annually.
- Maintain a documented risk management program aligned to NIST AI RMF or ISO/IEC 42001.
- Notify candidates when a high-risk AI system is used in a consequential decision, and offer an appeal and human review.
- Report any known algorithmic discrimination to the Colorado Attorney General within 90 days.
Enforcement sits with the Attorney General. There is no private right of action, but a documented impact assessment is an affirmative defense — which makes it the single most valuable artifact your firm can produce in 2026.
California: CCPA/CPRA and FEHA
California does not yet have an AEDT statute, but two existing frameworks already apply. Under CCPA/CPRA, candidates are consumers: they have a right to know what personal information you collect, how automated decision-making is used, and to request deletion. The CPPA's 2024 automated decision-making regulations tighten the disclosure and opt-out requirements. Under FEHA, the Civil Rights Department's proposed rules treat AI hiring tools as potential sources of disparate impact, with the employer or staffing agency liable regardless of vendor claims.
Other states to watch
- Maryland (HB 1202): Written candidate consent required before using facial recognition in a job interview.
- Washington (WAC 162-04A): Human Rights Commission guidance treats algorithmic disparate impact as a FEHA-equivalent violation.
- New York State AEDL: Pending legislation that would extend NYC-style bias audits statewide. Draft bills in 2025 and 2026 sessions.
- New Jersey, Massachusetts, Texas, Virginia: Active 2026 legislative proposals covering disclosure, bias testing, and impact assessments.
A practical 5-step compliance framework for US staffing firms
- Classify every tool. Is it an AEDT that substantially assists a decision, or an assistive drafting tool used by a recruiter? Claude Cowork used to summarize a resume for a human screener is assistive. An ATS that auto-ranks candidates is an AEDT. The distinction drives every downstream duty.
- Map your geography. Where does the candidate live? Where is the role performed? Where is your office? Local Law 144, AIVIA, SB24-205 and CCPA each pick a different anchor. A single placement can trigger two or three.
- Document candidate notice and consent. One plain-English paragraph in your application, interview invite and ATS footer. Store the timestamp with the candidate record.
- Commission a bias audit annually. Even where not strictly required. An audit is cheap insurance, satisfies NYC on the spot, and is the cornerstone of a Colorado impact assessment.
- Train your recruiters. Every state regime assumes a human in the loop who understands what the AI is doing. A documented training program (ours, or anyone's) is defense evidence.
This article is general information for staffing and recruiting firms. It is not legal advice. Consult qualified employment counsel in each state where you operate before adopting an AI hiring tool.