On September 30, 2026, California’s governor signed SB 947, which stops employers from disciplining or firing a worker on the word of an automated system alone. Its duties start on July 1, 2027. It is the newest of 13 enacted rules on AI in hiring, discipline and firing that we found across six states and New York City, and this page lists each one from the statute or the regulator’s own text.
- 01Firing is now coveredSB 947 is aimed at discipline and dismissal, not hiring, and applies to employers of every size.
- 02Colorado changed courseIts 2024 AI Act was delayed to June 30, 2026, then repealed and re-enacted as a narrower disclosure law that starts January 1, 2027.
- 03Connecticut is liveSince October 1, 2026, using an automated employment tool is no defence to a discrimination complaint.
- 04California stacks upCivil rights rules, privacy rules and SB 947 overlap, with three different start dates.
01 — The releaseWhat California signed on September 30 #
An automated decision system, or ADS, is any software that scores, ranks or recommends people, from a résumé screener to a productivity tracker. Under SB 947, chaptered as Chapter 859 of 2026, an employer may not rely only on such a system when deciding to discipline or dismiss someone. Where it relies mainly on the system’s output, a person has to check that output against other evidence, such as supervisor reviews, personnel files, the worker’s own output or witness interviews. If the output cannot be checked, or proves inaccurate, it cannot be used.
The employer must also give a written, plain-language notice at the time of the decision saying a system was mainly relied on and a person corroborated it. Workers can ask for a description of their own data the system used. The law has no size threshold and covers public employers and labour contractors. The bill states that it “shall become operative on July 1, 2027”.
Operative date
Set in §1526.7 of the bill. No duty applies before then.
Civil penalty
Plus injunctions, punitive damages and attorney’s fees in a civil action.
No size floor
Includes the state, cities, counties, school districts and labour contractors.
The governor signed three other workplace AI bills the same day. SB 951 requires a mass-layoff notice to say when AI or automation caused the layoff, under the heading “This notice is for a technology displacement”. AB 1883 bars AI surveillance tools that infer a worker’s emotions or collect neural data, with penalties of up to $500 per violation; it governs monitoring rather than hiring or firing, so it sits outside the table. AB 1331, on workplace surveillance, was not reviewed for this page.
02 — The dataThe table: what each law requires #
Rows run in the order the obligations start, so the table doubles as a timeline. Only enacted statutes, adopted regulations and local laws are included. Bills still in progress are covered in section five.
| Sources: statute, public act or regulator text for each row, read October 3, 2026. | |||
|---|---|---|---|
| Law | Starts | Covers | What it requires |
| --- | --- | --- | --- |
| New York City, Local Law 144 | Enforced from Jul 5, 2023 | Hiring and promotion screening by an automated tool | Independent bias audit no more than a year before use, with a summary posted; notice at least 10 business days ahead; candidates may ask for an alternative process or accommodation |
| Illinois, AI Video Interview Act | Jan 1, 2020 | AI analysis of recorded video interviews | Tell the applicant, explain what the AI assesses and get consent first; delete videos within 30 days of a request; report race and ethnicity data yearly if AI alone picks who gets an in-person interview |
| Maryland, Labor and Employment §3-717 | Oct 1, 2020 | Facial recognition during job interviews | No facial template from an interview without the applicant’s signed waiver |
| California, Civil Rights Council regulations | Oct 1, 2025 | Employment decisions under the state’s anti-discrimination law | Automated decisions that harm people on protected characteristics may be unlawful; keep records, automated-decision data included, for four years |
| Illinois, Human Rights Act amendment | Jan 1, 2026 | Recruitment through discipline, discharge and terms of work | Using AI with a discriminatory effect, or zip codes as a proxy, is a civil rights violation; so is failing to tell employees AI is used. Notice rules still pending |
| Texas, TRAIGA (HB 149) | Jan 1, 2026 | Any AI system, employers included | No developing or deploying AI with intent to discriminate; disparate impact alone does not prove intent. Its disclosure rules exclude employment |
| Connecticut, Public Act 26-15 §§13–14 | Oct 1, 2026 | Discrimination complaints | Using an automated employment decision tool is no defence; anti-bias testing may be weighed |
| Connecticut, Public Act 26-15 §26 | Oct 1, 2026 | Layoffs by employers filing a federal WARN notice | Tell the state Labor Department whether the layoffs relate to AI or other technological change |
| Colorado, SB 26-189 | Jan 1, 2027 | Employment and employment opportunities | Notice when a person deals with the system; a plain-language explanation within 30 days of an adverse outcome; a right to correct data and to human review where commercially reasonable |
| California, CPPA rules on automated decisions | Compliance by Jan 1, 2027 | Hiring, work and pay allocation, promotion, demotion, suspension, termination | Notice before use, an opt-out with exceptions (one is a human appeal), access to information and an appeal |
| California, SB 951 | Not stated in the text we read | Mass layoffs caused by AI or automation | The 60-day layoff notice must give numbers, jobs and sites affected, the functions automated and the kind of AI used |
| California, SB 947 | Jul 1, 2027 | Discipline and termination | No discipline or firing on an automated system alone; a human must corroborate output the employer mainly relies on; written notice with the decision |
| Connecticut, Public Act 26-15 §§7–12 | Oct 1, 2027 | Hiring, promotion, discipline, discharge, terms | Disclose the technology; give written notice before the decision with its purpose, trade name and the data it uses |
Two patterns stand out. The early laws, in New York City, Illinois and Maryland, target hiring tools and ask for consent, notice or an audit. The newer ones reach further into the job: Illinois covers discipline and discharge, California’s SB 947 is aimed squarely at them, and Connecticut’s Public Act 26-15 requires notice before any covered decision from October 2027.
03 — TimelineWhen each obligation starts #
- Illinois AI Video Interview ActConsent before AI reviews an interview
- Jan 1, 2020
- Maryland facial recognition ruleWaiver for interview face scans
- Oct 1, 2020
- New York City Local Law 144Enforcement began
- Jul 5, 2023
- California civil rights regulationsAutomated decisions under anti-discrimination law
- Oct 1, 2025
- Illinois Human Rights Act; Texas TRAIGABoth took effect the same day
- Jan 1, 2026
- Connecticut §13 and §26No defence; layoff disclosure
- Oct 1, 2026
- Colorado SB 26-189; California CPPA rulesDisclosure law; privacy-law compliance date
- Jan 1, 2027
- California SB 947Human corroboration for discipline and firing
- Jul 1, 2027
- Connecticut notice dutiesDisclosure and pre-decision notice
- Oct 1, 2027
04 — The dataWho enforces each law, and the penalties #
Penalties vary more than the duties do. Texas has the largest figures, but they need proof of intent to discriminate. New York City’s are small per violation but count each day a tool is used. Two name no penalty at all; several others rely on existing discrimination remedies.
| Sources: enforcement sections of each statute or regulation, read October 3, 2026. SB 951 is omitted: its text adds no new remedy. | |||
|---|---|---|---|
| Law | Who is covered | Enforcer | Penalty |
| --- | --- | --- | --- |
| New York City, Local Law 144 | Employers and agencies, for candidates and employees living in the city | Consumer and Worker Protection department | Up to $500 for a first violation, $500 to $1,500 for each later one; each day of use counts |
| Illinois, AI Video Interview Act | Employers hiring for Illinois-based roles | None named; data goes to the commerce department | None stated |
| Maryland, §3-717 | Employers | None named | None stated |
| California, Civil Rights Council regulations | Employers under the state’s anti-discrimination law | Civil Rights Department | Existing discrimination remedies |
| Illinois, Human Rights Act amendment | Employers under the Act | Department and Commission of Human Rights | Existing Human Rights Act remedies |
| Texas, TRAIGA | Anyone doing business in Texas | Attorney General only; 60 days to cure | $10,000 to $12,000 curable; $80,000 to $200,000 uncurable; $2,000 to $40,000 a day if continuing |
| Connecticut, §13 and §26 | Employers; federal WARN filers | Human rights commission and courts; Labor Department receives §26 disclosures | Existing discrimination remedies; none stated for §26 |
| Colorado, SB 26-189 | Developers and deployers doing business in Colorado | Attorney General only; 60 days to cure until 2030 | Consumer protection penalties; amount not confirmed |
| California, CPPA rules | Businesses covered by the state privacy law | Privacy Protection Agency and Attorney General | State privacy law penalties |
| California, SB 947 | All employers, public bodies and labour contractors included; no size floor | Labor Commissioner and public prosecutors | $500 per violation, plus injunctions, punitive damages and attorney’s fees in court |
| Connecticut, §§7–12 | Deployers doing business in Connecticut; no size floor | Attorney General only; 60 days to cure through 2027 | Unfair trade practices penalties |
New York City’s enforcement page sets out the audit and notice rules the city enforces. Texas’s law, Colorado’s and Connecticut’s notice duties create no private right of action and leave enforcement to the state attorney general.
05 — The catchThree claims that are now out of date #
- “The Colorado AI Act starts June 30, 2026.” That date is no longer the one to plan for. Lawmakers first pushed it to June 30, 2026, then passedSB 26-189 , signed on May 14, 2026, which repealed and re-enacted it as a narrower law on automated decisions. The new law, which drops the old impact-assessment and risk-programme duties, takes effect January 1, 2027.
- “SB 947 applies from January 1, 2027.” The bill is a non-urgency statute, but its own text sets the operative date at July 1, 2027.
- “Connecticut gives workers an explanation and a chance to correct their data.” Some legal summaries say so, but the enacted Public Act gives workers only a disclosure duty and a notice before the decision. Those explanation and correction rights are in Colorado’s law.
Bills still pending include a New York measure on statewide standards for automated hiring tools. A second New York bill, on reporting AI’s effect on hiring to the state labour department, passed both chambers according to legislative trackers; we could not confirm the governor’s action, so it is not in the table.
This page summarises enacted text for orientation. Coverage thresholds, exemptions and remedies turn on definitions we have shortened here. Check the full text, and the employment lawyer who knows your workforce, before changing a process.
06 — Practical implicationsWhat employers using AI should do now #
The common first step is an inventory: every tool that scores, ranks or flags people, which decisions it feeds and who reviews its output. Most of these laws ask some version of the same three questions, so one inventory serves all of them. Our guide to bias audits for smaller firms covers the audit side, and our HR team playbook covers where AI fits in recruiting. For the wider state-versus-federal picture, see our note on federal preemption. Teams that want the inventory and review process designed with them can start with our AI transformation service.
07 — MethodMethod and as-of date #
A table of enacted rules, read from primary legal text. Law-firm and tracker pages were used only to locate statutes.
- What was collected
- For each enacted statute, public act, adopted regulation or local law on AI or automated decisions in hiring, promotion, discipline or firing: start date, decisions covered, duties, who is covered, enforcer and penalty. 13 rows from 11 laws and regulations.
- Sources
- California legislative records and the Civil Rights Department and Privacy Protection Agency; New York City’s Administrative Code and its consumer and worker protection department; the Illinois, Colorado, Connecticut, Maryland and Texas legislatures.
- As-of date
- October 3, 2026. The newest rows are California SB 947 and SB 951, both chaptered September 30, 2026.
- Exclusions
- Bills not enacted; workplace surveillance laws such as California AB 1883; general privacy laws without an employment duty; federal guidance.
- Limitations
- SB 951’s operative date, Colorado’s penalty amount, the coverage thresholds of California’s privacy law and the Illinois notice-rule status were not confirmed from primary text.
- Refresh
- Re-checked monthly and when a covered legislature adjourns or a regulator issues rules. Changes are made in place with a dated note.
Inventory every system that scores your people
List each tool that ranks, flags or scores candidates and staff, the decisions it feeds and the person who reviews it. That one list answers most of what New York City, Colorado, Connecticut and California now ask, and it shows where human review is missing before July 2027.