Key takeaways
- Identify the jobs, candidates and decisions in scope before choosing the jurisdiction records to examine.
- Read each instrument’s scope, status and governing date together before planning implementation.
- Ask for the evidence and support the relevant duties need, including requirements that may apply during the contract.
New York City’s rules require a recent bias audit and candidate notice for covered hiring tools. Illinois’s employment AI provisions took effect in January 2026, while the rules specifying its notice process had not been adopted at this edition’s review date. A rollout plan needs to distinguish the duty, its timing and the process required to meet it.
Use this watch to distinguish a duty already in force from a future requirement, a proposal, or an enforcement restriction. A vendor’s description of a tool as compliant does not resolve those questions for the employer.
Status values used below: in force (duties apply now); enacted, not yet applying; proposed (not law); repealed; stayed (a court order bars enforcement). Each record identifies the governing instrument and the dated procedural information relied on. Recheck pending court orders and rulemaking before a rollout or renewal. This edition covers the jurisdictions below; other states, pending bills, the GDPR’s rules on automated decisions, biometric privacy laws such as Illinois’s and the federal Uniform Guidelines on selection procedures (covered in Interviewing for AI Judgment) are outside it. This page is orientation, not legal advice.
Explore the jurisdiction records
Jurisdiction navigator · editorial snapshot 20 October 2026
Start with geography. Read the scope before deciding what applies.
Select a highlighted state, the city inset or a regional record to reach its full discussion. The highlights identify records in this edition; they do not determine legal applicability.
Pan the state map or choose a jurisdiction from the links below.
City record
Enlarged borough outlines. New York City and New York State have separate records.
Regional record
European mainland focus with Cyprus and Malta. Overseas areas omitted.
Federal record
Alaska and Hawaii are insets at separate scales. Territories are not shown.
- Teal: a state, city or EU record is included.
- Gray: geographic context outside the selected records.
- Navy: the separate federal record.
This is a selected watch, not a complete inventory of employment law. Read each record’s scope, status and governing date together. Map boundaries are generalized for navigation, not legal or survey boundaries. Read the dated summaries below.
At a glance · this edition’s recorded position
Read the duty and its status together
Snapshot: 20 October 2026. These summaries reproduce the article’s recorded status; consult the full jurisdiction record and primary sources before use.
Record 01
New York City
Record summary
Bias audit, published results, advance notice
Status and governing date
In force since 2023
Record 02
Illinois
Record summary
No discriminatory-effect AI use; notice
Status and governing date
In force since January 1, 2026; notice rules not yet adopted
Record 03
Illinois (video interviews)
Record summary
Notice, explanation, consent, deletion
Status and governing date
In force since 2020
Record 04
California (civil rights rules)
Record summary
Existing discrimination law applied to automated-decision systems; four-year records
Status and governing date
In force since October 1, 2025
Record 05
California (privacy rules)
Record summary
Risk assessment before automated hiring decisions or inferring workers’ traits; pre-use notice, opt-out or human appeal, access
Status and governing date
Risk assessments since January 1, 2026 (by December 31, 2027 for uses begun earlier); notice, opt-out and access from January 1, 2027
Record 06
Colorado
Record summary
Notice, explanation after an adverse outcome, correction and human review
Status and governing date
Applies from January 1, 2027; enforcement stayed by a stipulated court order
Record 07
New Jersey
Record summary
Disparate-impact standard, including for vendors’ tools
Status and governing date
In force since December 15, 2025
Record 08
Connecticut
Record summary
AI use is no defense; pre-decision notice
Status and governing date
No-defense rule in force since October 1, 2026; notices from October 1, 2027
Record 09
New York State
Record summary
Public list of tools used by state agencies
Status and governing date
In force for state agencies only
Record 10
Texas
Record summary
Ban on intentional discrimination only
Status and governing date
In force since January 1, 2026
Record 11
European Union
Record summary
High-risk duties for employment AI; workplace emotion-recognition ban
Status and governing date
Ban in force; high-risk duties from December 2, 2027
Record 12
U.S. federal
Record summary
Existing discrimination statutes; earlier EEOC AI guidance removed
Status and governing date
Enforcement posture changed by executive order
Geography: U.S. Census Bureau 2024 cartographic boundaries; European country outlines Natural Earth 5.1.1, public domain. Boundary sources establish geography, not the status of the legal records.
What this means now
- Duties that apply today. Employers hiring in New York City need a current bias audit and candidate notice for covered tools. In Illinois, AI that has the effect of discriminating is unlawful whatever the intent. New Jersey prohibits practices with a disparate impact unless they are necessary to a substantial, legitimate, nondiscriminatory interest. In California, existing discrimination law already applies to automated-decision systems. California’s privacy rules have required a risk assessment since January 1, 2026 before a business starts using automated decisionmaking technology for hiring decisions.
- 2027 is when process duties arrive. California’s privacy rules require pre-use notice, opt-out and access for automated hiring decisions from January 1, 2027, and Colorado’s law applies from the same date, subject to its stay. Connecticut’s notices follow on October 1, 2027, and the EU’s high-risk duties for employment AI on December 2, 2027.
- Vendor contracts signed now will run into those dates. Ask now for the evidence and support those rules will require. The questions at the end of this page are a starting point.
New York City: bias audits and notice, with enforcement found ineffective
- Date that governs: January 1, 2023, when Local Law 144 took effect; enforcement of the law and its rules began July 5, 2023.
- Source tier: Official (New York City; New York State Comptroller).
- What it requires: An employer or employment agency in New York City may not use an automated employment decision tool to screen a candidate or employee unless the tool had a bias audit no more than one year before use and a summary of the results is public on its website. A covered tool is a computational process that issues a simplified output, such as a score or recommendation, used to substantially assist or replace discretionary decision making. The city’s rules define that as relying solely on the output, weighting it more than any other criterion, or using it to overrule human conclusions. Candidates who live in the city must get notice at least ten business days before use, and the city’s rules require the notice to explain how to request an alternative selection process or accommodation. The audit must at least calculate selection rates and impact ratios by sex, race and ethnicity, and their intersections. Under the law’s text, whose implementing rules took effect on July 5, 2023, civil penalties run up to $500 for a first violation and $500 to $1,500 for each later one, with each day of non-compliant use a separate violation.
- What changed: In December 2025 the New York State Comptroller reported on the city’s enforcement from July 2023 to June 2025. The city agency surveyed 32 companies and found one issue of non-compliance; the auditors, reviewing the same companies, found at least 17 potential instances. The agency received two complaints in two years. Of 12 test calls the auditors made to the city’s 311 line, three were correctly transferred. The agency disagreed with some findings but agreed to adopt recommendations.
- So what: Light enforcement does not change the duty. An employer that uses a tool in a way the rule covers, and has not commissioned an audit or posted notice, is out of compliance whether or not anyone has complained.
- Related reading: Interviewing for AI Judgment.
- Sources: New York City Council, Local Law 144 of 2021, December 11, 2021; New York City Department of Consumer and Worker Protection, Rules, 6 RCNY 5-300 to 5-304, automated employment decision tools, effective July 5, 2023; Office of the New York State Comptroller, “Department of Consumer and Worker Protection: Enforcement of Local Law 144 – Automated Employment Decision Tools,” Report 2024-N-6, December 2025.
Illinois: AI that has a discriminatory effect is a civil rights violation
- Date that governs: January 1, 2026, when Public Act 103-0804 took effect.
- Source tier: Official (Illinois).
- What it requires: It is a civil rights violation for an employer to use AI that has the effect of subjecting employees to discrimination on a protected basis in recruitment, hiring, promotion, discipline, discharge or other terms of employment, or to use zip codes as a proxy for a protected class. Failing to notify an employee that AI is used for those purposes is also a violation. The law reaches any employer with one or more employees in Illinois during 20 or more weeks of the year.
- What changed: The state’s Department of Human Rights must adopt rules on when and how notice is given. It proposed rules in May 2026 and withdrew them in June 2026. Its July 2026 agenda anticipates a new proposal around November 10, 2026. The withdrawn proposal is not an adopted notice rule; check the department’s current rulemaking record before relying on a notice procedure.
- Also in force: Illinois’s Artificial Intelligence Video Interview Act, in effect since January 1, 2020, requires an employer that uses AI to analyze applicants’ recorded video interviews to notify them, explain how the AI works and what it evaluates, and obtain consent first. Videos may be shared only with those needed to evaluate the applicant and must be deleted within 30 days of a request; an employer that relies solely on AI to decide who gets an in-person interview must report applicants’ race and ethnicity annually.
- So what: The Illinois standard turns on effect, not intent. The statute’s notice duty took effect with the rest of the law, but the rules that would define its timing and form do not exist yet; how employers should give notice in the meantime is a question for counsel.
- Sources: Illinois General Assembly, Public Act 103-0804, effective January 1, 2026; Illinois Secretary of State, “Illinois Register, Volume 50, Issue 26,” Department of Human Rights notice of withdrawal, June 26, 2026; Illinois Secretary of State, “Illinois Register, Volume 50, Issue 29,” Department of Human Rights regulatory agenda, July 17, 2026; Illinois General Assembly, Artificial Intelligence Video Interview Act, 820 ILCS 42, as amended, effective January 1, 2022.
California: civil rights rules in force since 2025, privacy rules phasing in from 2026
- Date that governs: October 1, 2025, for the Civil Rights Council’s employment rules. The privacy agency’s rules took effect January 1, 2026, and risk assessments are required from that date; compliance with its automated-decision notice, opt-out and access rules is required from January 1, 2027.
- Source tier: Official (California Civil Rights Council; California Privacy Protection Agency).
- What the civil rights rules require: It is unlawful to use an automated-decision system or selection criteria that discriminate against an applicant or employee on a protected basis, and evidence of anti-bias testing, or its absence, is relevant to a claim or defense. An employer’s agent, including one acting through an automated-decision system, is itself an employer under the rules. Employment records, including automated-decision system data, must be kept for four years. The rules apply to employers of five or more people, counting employees outside California.
- What the privacy rules require: A significant decision includes providing or denying employment and compensation. A risk assessment is required before a business uses automated decisionmaking technology to make such a decision, and before it uses automated processing to infer applicants’ or employees’ traits from systematic observation. That duty has applied since January 1, 2026 to uses begun on or after that date; for uses already under way then, the assessment is due by December 31, 2027. For assessments conducted in 2026 and 2027, a business must submit summary information to the agency by April 1, 2028, with an attestation under penalty of perjury by a member of executive management; the agency or the Attorney General can require the full assessment at any time. From January 1, 2027, a business that uses the technology to make such a decision must give a pre-use notice, let people opt out and answer requests for access to information about the use. The opt-out does not apply if the business offers an appeal to a human reviewer with authority to overturn the decision or, for hiring, uses the technology solely to assess ability to perform the work without unlawful discrimination.
- So what: The civil rights rules apply existing discrimination law to tools now; the privacy rules already require risk assessments and add notice, opt-out and access duties from 2027. Under the first, a vendor acting for the employer through an automated system is itself treated as an employer.
- Sources: California Civil Rights Council, Final text of employment regulations regarding automated-decision systems, June 2025; California Civil Rights Department, “Civil Rights Council Secures Approval for Regulations to Protect Against Employment Discrimination Related to Artificial Intelligence”, June 30, 2025; California Privacy Protection Agency, Text of Regulations: CCPA Updates, Cybersecurity Audits, Risk Assessments, Automated Decisionmaking Technology (ADMT), and Insurance Regulations, September 2025; California Privacy Protection Agency, CCPA Regulations, consolidated text, effective January 1, 2026; California Privacy Protection Agency, rulemaking record for the regulations, approved September 22, 2025.
Colorado: a replaced law that applies from 2027 and is under a court stay
- Date that governs: January 1, 2027, when SB 26-189 takes effect for decisions made on or after that date.
- Source tier: Official (Colorado General Assembly; U.S. District Court for the District of Colorado; Colorado Attorney General).
- What changed: SB 26-189 repeals and reenacts the Colorado AI Act of 2024 as a law on automated decision-making technology in consequential decisions, including employment. The earlier act’s effective date had been moved to June 30, 2026 before it was replaced.
- What it requires: Before using covered technology to materially influence a consequential decision, a deployer must give notice. After an adverse outcome, it must provide within 30 days a plain-language description of the decision and the technology’s role, and an explanation of the person’s rights. The person may ask to correct inaccurate data and for meaningful human review by someone with authority to change the decision, to the extent commercially reasonable. Records must be kept for three years. Only the Attorney General enforces, with a 60-day opportunity to cure until 2030.
- The stay: On April 27, 2026, in a suit by xAI against the Attorney General, the court ordered on a joint motion that the Attorney General not begin enforcement of the 2024 act or any legislation replacing or amending it enacted that session, until 14 days after the court rules on a preliminary-injunction motion xAI is to file after the implementing rules are adopted. The United States has intervened as a plaintiff. The Attorney General’s proposed rules are open for comment, with a public hearing on October 26, 2026. The Department of Law released a regulatory analysis on October 21, 2026; that filing is part of the rulemaking, not a final rule.
- So what: The duties are law, but when they can be enforced depends on the rules and the court. Check the stay’s scope and any later court orders against the governing statute before relying on an enforcement assumption. This page will recheck the rulemaking and the docket after the October 26 hearing.
- Sources: Colorado General Assembly, Senate Bill 26-189, Session Laws 2026, chapter 131, May 14, 2026; U.S. District Court for the District of Colorado, X.AI LLC v. Weiser, No. 1:26-cv-01515, docket, April 9, 2026; Colorado Department of Law, “Colorado Automated Decision-Making Technology & Chatbot Safety Rulemaking”, regulatory analysis released October 21, 2026.
New Jersey: disparate-impact rules that reach vendors’ tools
- Date that governs: December 15, 2025, the effective date of N.J.A.C. 13:16.
- Source tier: Official (New Jersey Division on Civil Rights).
- What it requires: An employment practice with a disparate impact on a protected class is prohibited unless it is necessary to a substantial, legitimate, nondiscriminatory interest, and may still be prohibited if a less discriminatory alternative would serve the same interest. Where a practice relies on a vendor’s products or systems, the employer must take reasonable steps to ensure they comply; on adoption the Division said reasonable steps are not a complete defense. A tool that screens applicants by schedule must include a way to request an accommodation.
- Sources: New Jersey Office of the Attorney General, Division on Civil Rights, Notice of Adoption, N.J.A.C. 13:16, Rules Pertaining to Disparate Impact Discrimination, December 15, 2025.
Connecticut: AI is no defense now, and notices from 2027
- Date that governs: October 1, 2026, for the no-defense provision; October 1, 2027, for the notice duties.
- Source tier: Official (Connecticut General Assembly).
- What it requires: Use of an automated employment-related decision technology is not a defense to a discrimination complaint, and evidence of anti-bias testing may be considered, including its quality, recency and results. From October 1, 2027, a deployer using such technology as a substantial factor in an employment decision must give the employee or applicant written notice before the decision, including the tool’s purpose, trade name and the data it analyzes. Developers must give deployers the information they need. Only the Attorney General enforces the notice sections, and there is no private right of action.
- Sources: Connecticut General Assembly, Public Act 26-15, An Act Concerning Online Safety, May 27, 2026; Connecticut General Assembly, Public Act 26-100, June 2, 2026.
New York State and Texas: narrower rules
- New York State: Chapter 96 of 2025 requires state agencies that use automated employment decision tools to publish a list of them each year, and bars their use of AI from displacing state employees; that protection expires July 1, 2028. It does not apply to private employers.
- Texas: TRAIGA, in force since January 1, 2026, bars developing or deploying AI with the intent to discriminate unlawfully, states that disparate impact alone does not show intent, and contains no employment-specific duties. Its definition of a consumer excludes people acting in an employment context.
- Sources: New York State Legislature, S822, chapter 96 of 2025, February 14, 2025; Texas Legislature, H.B. 149, Texas Responsible Artificial Intelligence Governance Act, effective January 1, 2026.
European Union: employment AI is high-risk from December 2027; emotion recognition at work is already banned
- Date that governs: December 2, 2027, when the AI Act’s high-risk obligations apply to employment systems after the Digital Omnibus; the original date was August 2, 2026.
- Source tier: Official (European Union).
- What it covers: AI intended for recruitment or selection, including filtering applications and evaluating candidates, and AI used for decisions on promotion or termination, task allocation, or monitoring and evaluating performance.
- What deployers will owe: Employers must inform workers’ representatives and affected workers before using a high-risk system at work. Deployers must inform people that they are subject to the system and assign human oversight to people with the necessary competence, training and authority. Logs must be kept for at least six months. People affected by an adverse decision have a right to a clear explanation of the system’s role. When that right applies to employment systems is not settled: the Omnibus moved the application dates for Sections 1 to 3 of the high-risk chapter, and the explanation right sits in a different chapter of the Act.
- Already in force: Using AI to infer the emotions of people in the workplace is prohibited, except for medical or safety reasons.
- Also changed: The Omnibus replaced the AI literacy duty; employers must now take measures to support their staff’s AI literacy, without having to guarantee any level. The amending regulation was published on July 24, 2026.
- Sources: European Union, Regulation (EU) 2024/1689 (Artificial Intelligence Act), June 13, 2024; European Union, Regulation (EU) 2026/1744 (Digital Omnibus on AI), July 8, 2026.
United States: existing discrimination law, with earlier AI guidance removed
- Source tier: Official (EEOC; Executive Office of the President; Department of Labor).
- What changed: EEOC removed its 2022 and 2023 technical assistance on AI in employment assessments from its website in late January 2025. An executive order of April 2025 makes it U.S. policy to eliminate disparate-impact liability to the maximum degree possible and directs agencies to deprioritize its enforcement, naming Title VII. The Labor Department’s AI and worker well-being principles page no longer loads. For the changes to the federal selection guidelines, see Interviewing for AI Judgment.
- So what: These changes come from the executive branch and do not amend Title VII, the ADA or the ADEA. State law and private suits continue regardless.
Litigation: Mobley v. Workday
- What it is: An employment discrimination case filed in February 2023 in federal court in California, alleging discrimination through an AI screening vendor’s tools.
- Where it stands: In July 2024 the court held that the complaint adequately alleged that Workday acted as an agent of its client-employers, and so could be treated as an employer under federal discrimination law; these are rulings on allegations, not findings. In May 2025 it preliminarily certified a collective on the age discrimination claim. In June 2026 it dismissed some claims and let others proceed, and in July 2026 it declined to certify that ruling for immediate appeal. A motion for class certification was filed on September 14, 2026. The discovery order filed October 7, 2026 recorded November 10 as the deadline for Workday’s opposition to class certification. Discovery was continuing in October 2026.
- So what: If the agency theory holds, a vendor’s screening tool can create liability for the vendor as well as the employer. The case does not yet show how a court will decide either.
- Sources: U.S. District Court for the Northern District of California, Mobley v. Workday, Inc., No. 3:23-cv-00770-RFL, docket, February 21, 2023; U.S. District Court for the Northern District of California, Mobley v. Workday, Inc., order granting in part and denying in part motion to dismiss, ECF 360, June 22, 2026; discovery order, ECF 435, filed October 7, 2026.
What to ask your vendors
Several of these rules reach tools a vendor builds or runs. Before relying on a vendor’s AI in hiring, ask for:
- The tool’s role in the decision: does it score, rank, screen out or only inform, and can a person override it?
- Any bias audit or anti-bias testing, with its date, method and results by group.
- What notice and explanation the vendor supports for candidates, and in which jurisdictions.
- What data the tool uses, how long it is kept, and how candidates can correct it.
- Who is responsible for compliance under the contract, and what the vendor will provide if a candidate, regulator or court asks.
Review status
Orientation, not legal advice. Several statuses will change before the next scheduled review: the Colorado hearing on October 26, 2026, the expected Illinois proposal in November 2026, and the Workday filings due November 10, 2026.


