Hiring Resources
12 min Read

AI Hiring Laws 2027: Employer Compliance Guide for NYC, Illinois, Colorado, California and the EU

Mayank Pratap Singh
Mayank Pratap Singh
Co-founder & CEO of Supersourcing

Three AI hiring laws already bind US employers today, and two more land in six weeks. Illinois has required candidate notice for AI-assisted hiring decisions since 1 January 2026, New York City has required bias audits since 2023, and California’s fair-employment regulations have treated automated screening tools as the employer’s conduct since October 2025. 

California’s automated decision-making rules and Colorado’s rewritten AI statute follow on 1 January 2027, and in every one of the five the obligation falls on the employer, not on the vendor whose software did the screening. 

That is the part most compliance plans miss: the résumé ranker lives in a staffing agency’s stack, the employer never configured it, and the notice, disclosure and record-keeping duties still land on the employer. 

Courts have been moving the same way. In Mobley v. Workday the Northern District of California has let claims proceed on the theory that a screening vendor can be the employer’s agent, held in March 2026 that age-discrimination disparate-impact protection reaches applicants, and in June 2026 allowed California’s fair-employment law to reach the tool even for jobs outside the state. 

The AI hiring laws that apply to a US employer in 2027 are manageable if they are treated as a procurement problem rather than a legal abstraction: know which tools touch your candidates, who runs them, what each jurisdiction requires, and what the contract with your staffing vendor says about all three. 

This guide covers each regime with dates and obligations, the federal and litigation picture, the five clauses that belong in every staffing agreement, and a ten-question audit for any vendor that screens on your behalf.

It is written for HR, legal and procurement leaders and for the engineering managers who sign off on hiring tools. It is not legal advice; it is the checklist that makes your counsel’s review shorter.

The EU’s Digital AI Omnibus, in force since 27 July 2026, moved the AI Act’s high-risk obligations for employment and recruitment systems from 2 August 2026 to 2 December 2027, while the Article 5 prohibitions still apply from 2 December 2026. 

TL;DR

This guide is for HR, legal, procurement and engineering leaders who use, or whose staffing vendors use, AI to screen candidates. It sets out the AI hiring laws in force or taking effect across NYC, Illinois, Colorado, California, Texas and the EU, with dates, and turns them into an AI in hiring compliance checklist for employers and their vendors.

The biggest point: in every one of these regimes the employer carries the notice, disclosure or audit duty even when a staffing agency or ATS vendor runs the tool, and in Mobley v. Workday the vendor is being treated as the employer’s agent.

By the end you will know which AI hiring laws apply to each of your hiring locations, the five clauses to add to a staffing MSA, and the ten questions that tell you whether a vendor’s screening tool is a compliance asset or a liability.

 

What Are AI Hiring Laws?

AI hiring laws are statutes and regulations governing the use of automated or AI-driven tools to screen, rank, interview or select candidates and employees. They typically require candidate notice, bias audits or impact assessments, disclosure and human review when AI materially influences a decision, and record-keeping, binding the employer whether the tool runs in-house or at a vendor.

The “whether the tool runs in-house or at a vendor” clause is the operative part for anyone who buys IT staffing. The AI hiring laws differ on what they require; they agree on who is responsible.

The Core Problem: Your Vendor Runs The Tool, You Carry The Duty

Three legal threads converge on the employer. First, the AI hiring laws name the employer or “deployer” as the duty-holder: NYC’s bias-audit and notice rules, Illinois’s notice requirement, Colorado’s disclosure and record-keeping duties and California’s rules all attach to the entity making the employment decision, not to the software company. Second, California’s fair-employment regulations, effective 1 October 2025, define an automated decision system to include tools that merely “facilitate” a decision, and let agencies weigh the presence or absence of anti-bias testing in a discrimination claim. Third, Mobley v. Workday has advanced the theory that the vendor is the employer’s agent, which makes the vendor’s tool the employer’s conduct.

The practical consequence is a gap most buyers have not closed, and it is where AI hiring laws meet procurement. In the staffing RFPs and agreements the Supersourcing delivery team sees, fewer than one in five ask which automated tools touch candidates, and almost none ask for an audit date. Yet every modern staffing vendor ranks, filters or scores with software, and the employer who cannot name the tool cannot give the notice, attach the audit or answer the 30-day disclosure request that Colorado will require from January 2027.

AI Hiring Laws By Jurisdiction: What Each One Requires In 2027

This section covers the six AI hiring laws a US employer with multi-state or EU hiring is most likely to meet, in the order they took or take effect. Each entry gives the scope, the employer obligations and the date, with a summary table for the compliance file. Dates are as of November 2026; the federal section explains why some may still move.

NYC Local Law 144: bias audits and candidate notice

In force since July 2023, the oldest of the US AI hiring laws covers automated employment decision tools used to substantially assist hiring or promotion decisions for NYC jobs. Employers must obtain an independent bias audit within the past year, publish a summary of its results, and give candidates at least ten business days’ notice that a tool will be used, with the job qualifications it assesses and a route to request an alternative process. 

The audit must be of the tool as used, so a vendor’s generic audit may not satisfy the bias audit requirements if your job data differs materially. LL144 remains the strictest of the AI hiring laws on auditing.

Illinois HB 3773: notice, and discriminatory effect is enough

Effective 1 January 2026, the amendment to the Illinois Human Rights Act makes it a civil-rights violation to use AI in recruitment, hiring, promotion, discipline, discharge or training in a way that has a discriminatory effect, and bars the use of ZIP codes as a proxy for protected class. Employers must notify applicants and employees when AI is used for those decisions. 

The Illinois Department of Human Rights published proposed notice rules on 15 May 2026 and withdrew them on 2 June 2026, so as of November 2026 the notice duty is in force without implementing regulations; written notice at application and before any AI-assisted decision is the conservative reading of the Illinois AI hiring law.

Colorado SB 26-189: the rewritten act, effective 1 January 2027

Colorado’s original AI Act never took effect. After xAI sued on 9 April 2026 and the Department of Justice intervened, a federal magistrate stayed enforcement on 27 April 2026, and the legislature repealed and replaced the statute with SB 26-189, signed 14 May 2026 . The replacement drops the risk-management programme and annual impact assessment. 

From 1 January 2027, employers using automated decision-making technology that materially influences an employment decision must give public notice of use, disclose within 30 days of an adverse decision that automation was involved, what role it played and how to seek reconsideration, honour access, correction and human-review requests, keep records for three years, and confirm the vendor has supplied the required developer documentation. 

The Colorado AI Act for employers is therefore lighter than the 2024 version but still the only US regime with a statutory 30-day adverse-decision disclosure; enforcement waits on attorney-general rulemaking and the stay in xAI v. Weiser.

California: FEHA regulations now, CCPA automated-decision rules from 1 January 2027

Two layers of AI hiring laws apply in California. The Civil Rights Council’s FEHA regulations, effective 1 October 2025, treat automated decision systems as subject to the state’s anti-discrimination law and make anti-bias testing relevant evidence. 

The California Privacy Protection Agency’s regulations on automated decision-making technology, finalised in 2025, take effect 1 January 2027 for significant decisions including hiring, termination and compensation: pre-use notice, opt-out or human-appeal rights in defined cases, risk assessments and access rights. 

Because Mobley held that a vendor’s California nexus can bring out-of-state jobs within FEHA, employers outside California that use California-based screening vendors should treat these rules as applying to them.

Texas TRAIGA: intent, not impact

The Texas Responsible Artificial Intelligence Governance Act, effective 1 January 2026, prohibits using AI in employment decisions with intent to discriminate. It is the narrowest of the six: a disparate-impact showing is not enough, and there is no audit or notice mandate. It matters mainly because it signals that the state AI hiring laws are not converging on one model.

EU AI Act: high-risk employment systems from 2 December 2027

Recruitment, selection, evaluation and promotion tools are high-risk under Annex III. The Digital AI Omnibus, adopted 29 June 2026 and in force from 27 July 2026, moved the compliance date for those systems from 2 August 2026 to 2 December 2027; the prohibitions, including emotion recognition in the workplace, apply from 2 December 2026, and transparency duties from 2 August 2026. 

Deployers of high-risk systems will owe human oversight, logging, worker information and fundamental-rights impact assessments; processing special-category data for bias testing is allowed only where strictly necessary. US employers hiring into EU entities, or using EU-based vendors, should plan to the 2027 date now; the EU rules are the most prescriptive of all the AI hiring laws on logging and oversight.

Jurisdiction Scope Employer obligations Date
New York City (LL144) Automated tools substantially assisting hiring or promotion for NYC roles Annual independent bias audit, published summary, 10 business days’ candidate notice, alternative process on request In force since July 2023
Illinois (HB 3773) AI in recruitment, hiring, promotion, discipline, discharge, training Notice to applicants and employees; no discriminatory effect; no ZIP-code proxies 1 January 2026 (rules withdrawn June 2026)
Colorado (SB 26-189) Automated decision-making technology materially influencing employment decisions Public notice, 30-day adverse-decision disclosure, human review, access and correction, 3-year records, vendor documentation 1 January 2027 (enforcement stayed pending rulemaking)
California (FEHA regs; CCPA ADMT) Automated decision systems in employment; significant decisions Anti-bias testing as evidence (now); pre-use notice, risk assessment, opt-out or appeal, access (2027) 1 October 2025; 1 January 2027
Texas (TRAIGA) AI used with intent to discriminate in employment Prohibition only; no audit or notice mandate 1 January 2026
EU (AI Act, Annex III) Recruitment, selection, evaluation, promotion systems Human oversight, logging, worker information, impact assessment; prohibitions on emotion recognition at work 2 December 2027 (prohibitions 2 December 2026)

The Federal Picture: Preemption Talk, No Preemption

Executive Order 14365, signed 11 December 2025, directed federal agencies to challenge, preempt or defund state AI statutes, including the state AI hiring laws above, and the Attorney General established an AI Litigation Task Force on 9 January 2026. As of 1 September 2026 the Task Force had not sued any state; the Department of Justice’s one intervention was in xAI v. Weiser against Colorado, and a March 2026 White House framework seeking broad preemption is non-binding. 

The litigation that actually affects employers is private: Mobley v. Workday (nationwide age collective preliminarily certified May 2025, motion to dismiss largely denied 22 June 2026) and Kistler v. Eightfold AI (filed 20 January 2026, alleging an AI screening vendor is an unregistered consumer reporting agency under the FCRA).

What that means for a 2027 plan: do not budget on preemption of the state AI hiring laws. Illinois, NYC and California’s FEHA rules are in force now; California’s 2027 rules and Colorado’s replacement are on the calendar; and the FCRA theory in Kistler, if it succeeds, would add adverse-action notice duties that no state law currently imposes. The IT staffing trends 2027 analysis treats this compliance stack as one of the ten structural shifts for the year, and it is the one most often left out of hiring budgets; the RPO ensures compliance guide covers how an outsourced recruiting function absorbs that work.

Are Staffing Agencies Covered By AI Hiring Laws? Yes, And So Are You

A staffing vendor that ranks or filters candidates with software is using an automated employment decision tool on your behalf, and the AI hiring laws treat it that way. Under the agent theory in Mobley, the FEHA “facilitate” definition and Colorado’s deployer duties, the employer remains responsible for notice, disclosure and records even when it never configured the tool. 

The fix is contractual, whether the vendor’s software is a résumé ranker or one of the AI in recruitment tools sold to employers directly. These five clauses belong in every staffing MSA, as the compliance exhibit the IT staffing agreement guide describes:

  1. Tool disclosure. A named list of every automated or AI tool used on your candidates, updated within 30 days of any change, with the vendor and the decision step each tool touches.
  2. Audit attachment. For each tool, the most recent bias audit or impact assessment, its date and methodology, and an obligation to refresh annually or on material change; for NYC roles, an audit of the tool as used on your job data.
  3. Notice duty. Who gives the candidate notice (vendor or employer), in what form, how many days before use, and the template text; for Illinois and NYC roles, at application.
  4. Data handling. Consent basis, retention period, deletion on request and the jurisdictions the data moves through, including DPDP in India for offshore sourcing (AZB & Partners, 2026) and GDPR for EU candidates.
  5. Indemnity and records. Vendor indemnity for claims arising from its tools, an obligation to preserve tool logs and outputs for at least three years, and cooperation in any 30-day disclosure or agency inquiry.

The US IT staffing compliance guide covers the classification and co-employment clauses that sit next to these; the two sets belong in the same exhibit.

The Ten-Question Vendor AI Audit

Ask these in the RFP or at the next quarterly review. A vendor that can answer all ten in writing under the AI hiring laws for your locations is a compliance asset; one that cannot answer the first three is using tools it has not inventoried.

  1. Which automated or AI tools touch our candidates, at which step, and who supplies each one?
  2. When was each tool last bias-audited, by whom, on what data, and may we see the report?
  3. Does any tool make or substantially assist a screening decision without human review, and if so where?
  4. What notice do candidates receive, when, and in which jurisdictions?
  5. How do you handle a candidate who requests an alternative process or human review?
  6. What data does each tool use, where is it stored, for how long, and under which consent basis?
  7. Have you tested for disparate impact by age, disability, race, sex and ZIP-code proxies, and how does your process compare with the ways RPO mitigates hiring bias?
  8. What is your process when a tool is updated or replaced mid-contract?
  9. Which of our hiring locations have you mapped to NYC, Illinois, Colorado, California, Texas and EU obligations?
  10. Will you indemnify us for claims arising from your tools, and preserve logs for three years?

The IT staffing RFP guide carries these as questions 14 and 21–25 of its vendor annex; the IT staffing agency vetting process shows what a stage-by-stage answer to question 3 looks like when humans run the technical rounds. Pair the questionnaire with the background checks for remote employees steps so screening and verification sit in one file.

Two Engagements, Anonymised

US SaaS, multi-state hiring with three staffing vendors. A procurement review against the AI hiring laws for its hiring states found none of the three agreements named a screening tool; two vendors ran résumé-ranking software and one used an automated video screen for NYC roles with no audit on file. The client added the five clauses at renewal, dropped the vendor that could not produce an audit, and had notices, audit attachments and a tool inventory in place eight weeks before the 2027 dates. No rate changed; one supplier did.

Indian GCC, EU and US parent. The GCC’s sourcing partner ran AI ranking on candidates for roles reporting into EU and US entities. The agreement was amended to add tool disclosure, DPDP consent and retention terms, human review before any rejection, and an EU-ready logging obligation ahead of the 2027 Annex III date. The sourcing-to-shortlist timeline stayed at 7–10 working days because the screening stages were already human-run; the change was in documentation, not process, which is the usual shape of AI hiring laws work when vetting is already done by people.

What A Compliant Staffing Partner Looks Like

What to check Compliant partner Exposure
Tool inventory Named list, updated within 30 days, provided unprompted “We use AI-powered sourcing” with no tool named
Bias testing Dated audit per tool, methodology stated, refreshed annually “Our vendor handles that”
Human review Engineers run technical rounds; no automated rejection without review Automated score thresholds reject candidates silently
Notice and disclosure Jurisdiction-mapped notice templates; 30-day disclosure support Candidate notice left to the employer by default
Data and records Consent basis, retention, DPDP and GDPR terms, 3-year log retention Candidate data retained indefinitely, no consent record

A partner in the left column costs no more per seat than one in the right, and satisfies the AI hiring laws on your behalf; the difference is whether the compliance work is done once by the vendor or per hire by you. The IT staffing company in USA terms and the IT staffing company in India terms both carry the tool-disclosure and data exhibits, and the RPO compliance and regulations guide covers the same obligations for outsourced recruiting.

What Most Teams Get Wrong

They treat AI hiring laws as a software question and ask the ATS vendor for a compliance letter. The duty sits with the employer, the tools that matter are often the staffing vendor’s rather than the ATS, and no vendor letter satisfies a notice, disclosure or record-keeping obligation. Inventory every tool that touches candidates, including your vendors’, put the five clauses in every staffing agreement, and keep the audit reports in your file, not theirs.

Pressure-Test Your Vendors Before The January Dates

If you have staffing vendors screening candidates in New York, Illinois, Colorado, California or the EU and no tool inventory on file, send the vendor list and your hiring locations to mayank@engineerbabu.com. 

Supersourcing will return the ten-question audit completed for its own sourcing and vetting stack, the five-clause compliance exhibit used in its agreements, and a jurisdiction map for the roles you are hiring, so the AI hiring laws above become a line in the contract rather than a finding in an inquiry. 

The recruitment process outsourcing and hiring dedicated development teams pages describe the engagement models the exhibit attaches to, and the contact form is at https://supersourcing.com/contact-us/.

FAQ

What are the laws on AI in hiring?

The main US AI hiring laws are New York City’s Local Law 144, Illinois’s HB 3773, Colorado’s SB 26-189, California’s FEHA and CCPA automated-decision rules and Texas’s TRAIGA, alongside federal anti-discrimination law applied through cases such as Mobley v. Workday. In the EU, the AI Act treats recruitment and selection systems as high-risk from 2 December 2027. There is no federal AI hiring statute.

Do employers need a bias audit for AI hiring tools?

In New York City, yes: an independent audit within the past year, with a published summary, for any automated tool substantially assisting hiring or promotion decisions. Elsewhere the audit is not mandated by statute but is evidence California agencies may weigh, is the practical basis for Colorado’s vendor-documentation duty, and is the first thing a plaintiff requests in discovery.

Is it legal to use AI to screen resumes ?

Yes, in every US jurisdiction and the EU, subject to the conditions the AI hiring laws set: candidate notice where required, no discriminatory effect, human review and disclosure where a tool materially influences an adverse decision, and record-keeping. The tool must also respect existing anti-discrimination law, which is the basis of the Mobley claims against automated screening.

What is the Colorado AI Act?

Colorado’s 2024 AI Act was repealed and replaced in May 2026 by SB 26-189 after a federal court stayed enforcement of the original. From 1 January 2027 it requires employers using automated decision-making technology in employment decisions to give notice, disclose within 30 days when automation materially contributed to an adverse decision, allow human review, and keep records for three years. Enforcement awaits attorney-general rulemaking.

Who is liable when a staffing vendor uses AI screening?

Both parties can be. The AI hiring laws place notice, disclosure and audit duties on the employer as deployer; courts in Mobley v. Workday have allowed claims against the vendor as the employer’s agent. The staffing agreement should therefore name the tools, attach audits, allocate the notice duty, set data terms and carry a vendor indemnity, so liability and the records to defend it sit together.

How do we audit our staffing vendors for AI compliance before 2027?

Send the ten questions above to every vendor now, require written answers inside 30 days, and amend agreements at the next renewal with the five clauses, mapped to the AI hiring laws for each hiring location. For the questionnaire pre-filled for contract IT staffing, onsite developers or an enterprise RPO engagement, with tool inventory and audit dates attached, the address below returns it as a document rather than a pitch.

Author

  • Mayank Pratap Singh - Co-founder & CEO of Supersourcing

    With over 11 years of experience, he has played a pivotal role in helping 70+ startups get into Y Combinator, guiding them through their scaling journey with strategic hiring and technology solutions. His expertise spans engineering, product development, marketing, and talent acquisition, making him a trusted advisor for fast-growing startups. Driven by innovation and a deep understanding of the startup ecosystem, Mayank continues to connect visionary companies and world-class tech talent.

    View all posts

Related posts