Most recruiting teams now use some form of AI. Sourcing tools that rank profiles, screening tools that score resumes, chat assistants that ask knockout questions, interview tools that summarise answers. A few years ago none of this was regulated directly. That has changed. Several places now have rules aimed at AI in hiring, and more are coming. This guide covers the ones recruiting teams and agencies ask us about most, what each one actually requires and a short list of things to do now. It is a practical summary, not legal advice.
What these laws actually regulate
None of these laws ban AI in recruiting. They target a specific use: software that makes, or meaningfully influences, a decision about a person. Screening someone out, ranking a shortlist, scoring an interview. AI that drafts an outreach email, cleans up a job description or books an interview slot is usually outside that core scope.
That distinction is the most useful thing to understand. The more your AI acts as a decision maker about candidates, the more these rules apply. The more it acts as an assistant doing admin work while a recruiter makes the call, the less exposure you have. It is also why we design recruiting agents so a person makes every screen-out decision. More on that in our explainer on what AI recruiting agents should and should not do.
New York City: Local Law 144
New York City was first. Local Law 144 covers automated employment decision tools, or AEDTs, used by employers and employment agencies for candidates or employees in the city. Enforcement by the Department of Consumer and Worker Protection began on July 5, 2023. Before using an AEDT you need three things.
- A bias audit within the last year. An independent auditor checks selection rates and impact ratios across sex and race or ethnicity categories. You can rely on an audit for one year from the date it was done.
- A public summary. The summary of the most recent audit and the date you started using the tool go on the employment section of your website.
- Notice to candidates. Candidates must be told at least 10 business days before the tool is used, including how to request an alternative process or accommodation.
Penalties start at $500 for a first violation and go up to $1,500 for each later one. Two points catch teams out. First, agencies are covered in their own right, so a staffing firm screening for a client in New York cannot assume the client handles it. Second, a vendor telling you its tool is audited does not finish the job. You still need to confirm the audit exists, publish the summary and send the notice.
Illinois: HB 3773 and the AI Video Interview Act
Illinois has had the Artificial Intelligence Video Interview Act since 2020. It covers employers that use AI to analyse recorded video interviews and requires notice, an explanation of how the AI works and consent.
The bigger change is HB 3773, which amended the Illinois Human Rights Act and took effect on January 1, 2026. It makes it a civil rights violation to use AI in hiring, promotion and other employment decisions in a way that discriminates against protected groups. It also requires employers to give notice when AI is used to influence or facilitate those decisions. Draft rules from the Illinois Department of Human Rights apply to employers and their agents, which includes recruiters and other third parties acting for them. Under the draft rules, notice to applicants goes in the job posting, and AI related notices and disclosures are kept for four years.
Colorado: a new law from January 2027
Colorado passed the first broad state AI law in 2024, covering high-risk AI used in consequential decisions such as hiring. Its start date moved from February 2026 to June 30, 2026, and enforcement was then paused by a court order in a constitutional challenge. In May 2026 the state signed SB 26-189, which repeals and replaces that law from January 1, 2027. The new version focuses on notice, explanation and human review when AI materially influences a consequential decision. If you hire in Colorado, plan for those three duties to apply from the start of 2027.
The EU AI Act
For teams hiring in Europe, the EU AI Act lists AI used in recruitment and employment as high-risk in Annex III. That brings duties around risk management, data quality, logging, transparency and human oversight. The high-risk requirements were due on August 2, 2026, but the AI Omnibus changes deferred them. For standalone Annex III uses like hiring, they now apply from December 2, 2027. That is more time, not an exemption.
A practical checklist for recruiting teams
The rules differ in detail, but they push in the same direction. These steps cover most of the ground in every jurisdiction above.
- Inventory every AI tool in your hiring flow. Sourcing, screening, assessments, chat, interview analysis and ATS features. Note what each one does and where candidates are located.
- Mark which tools influence decisions. Anything that scores, ranks, filters or rejects people goes on a short list. Those are the tools the laws care about.
- Keep a human on screen-out decisions. The AI can prepare a shortlist with reasons. A recruiter approves or rejects. Log who made the call.
- Ask vendors for their audit and documentation. Request the latest bias audit, what data the tool uses and how it reaches a score. If they cannot answer, treat that as a risk.
- Write and send notices. Add AI use language to job postings and candidate emails where required, with the NYC 10 business day lead time if you hire there.
- Keep records. Notices, audits, decision logs and tool versions. Illinois draft rules point to four years, which is a sensible default.
- Review every year. Tools change, laws change and the NYC audit expires after a year.
What this means when you build AI recruiting agents
Compliance is easier when it is designed in rather than added later. When we build agents for recruiting teams, sourcing and outreach agents do the volume work, screening agents prepare a shortlist with written reasons, and a recruiter makes every advance or reject decision inside the ATS. Every step is logged with what the agent saw and what the person decided. Notices and opt-outs live in the same workflow, so nobody has to remember them.
If you want to see which parts of your process are admin work and which are decisions, start with our map of where recruiter hours go. Or read how our AI recruiting agents work inside the tools you already use.