Automated employment decision tools are software that scores, ranks or screens job candidates and employees. In 2026, New York City, California, Illinois, Ontario and, from January 2027, Colorado each regulate them differently, so a retailer hiring across regions needs one inventory, one bias-testing routine and one notice process that satisfies the strictest rule.
A retailer does not buy one hiring tool. It buys a résumé screener for corporate roles, a chat-based assistant for store applicants, a scheduling tool for seasonal staff and an assessment platform for managers. Each one can count as an automated employment decision tool, and each one sits under a different mix of city, state and provincial rules. A missed notice or a missing audit is a violation, and in New York City every day of use without the required notice counts separately.
What counts as an automated employment decision tool?
The label varies by statute, but the test is the same: does software produce a score, ranking or recommendation that a person then uses, or lets it replace the person, in a hiring or promotion decision? New York City’s Department of Consumer and Worker Protection (DCWP) explains that employers and employment agencies must bias-audit the tool within one year of use, post a summary of the results, and give candidates notice at least 10 business days before use. California goes wider and covers any “computational process that makes a decision or facilitates human decision making regarding an employment benefit.”
For a retailer, that wide definition pulls in tools HR may not think of as “AI”: the applicant-tracking system’s match score, a video-interview analyzer, a game-based assessment, and the vendor’s “recommended candidates” list.
Which laws apply to a retail employer in 2026?
| Jurisdiction | What it requires | Status |
|---|---|---|
| New York City (Local Law 144) | Annual independent bias audit, public summary, 10 business days’ notice | Enforced since July 5, 2023 |
| California (Civil Rights Council rules) | Discrimination rules extended to automated-decision systems; records kept four years | Effective October 1, 2025 |
| Illinois (HB 3773) | Notice when AI is used in employment decisions; no zip codes as a proxy for protected classes | Effective January 1, 2026 |
| Colorado (SB 26-189) | Pre-use notice; explanation within 30 days of an adverse decision; human review on request; three-year records | Effective January 1, 2027 |
| Ontario (job-posting rule) | Disclose AI screening in public job postings | In force since January 1, 2026 |
Treat the table as a map, not legal advice. Verify the current text of each rule before you act.
What does New York City require, and what does it cost to miss?
Local Law 144 asks for three things: an independent bias audit, a public summary of the results, and candidate notice. The audit tests the tool’s disparate impact on categories such as sex, race and ethnicity using an impact ratio. Penalties are $500 for a first violation and the same for each additional violation on that day, then $500 to $1,500 for each subsequent violation, and each day of non-compliant use counts as its own violation.
Take a retailer that posts 200 store openings before the holiday season and runs every applicant through one screening tool without notice. Under the daily-violation reading, the exposure builds with every day the tool stays live, not with every applicant. For the full NYC mechanics, see Fruggr’s NYC Local Law No. 144 page and the AI regulations overview below.
What changed in California, Illinois and Colorado?
California’s Civil Rights Council rules took effect on October 1, 2025. They require employers to keep employment records, including automated-decision data, for at least four years, up from two. Liability can reach an employer’s agent, which includes a recruiting vendor acting on the employer’s behalf. Retailers that outsource high-volume screening should read their vendor contracts with that in mind.
Illinois HB 3773 has applied since January 1, 2026. It requires notice when AI is used in covered employment decisions and bars the use of zip codes as a proxy for protected classes. The Illinois Department of Human Rights postponed its rulemaking on June 2, 2026, with no new timeline, so the statute is law while the detailed notice rules are not final.
Colorado replaced its 2024 AI Act with SB 26-189, signed on May 14, 2026. The Attorney General held its formal rulemaking hearing on October 26, 2026 and must adopt rules before January 1, 2027, when both the ADMT law and the rules take effect. Deployers must give notice before use and a plain-language explanation within 30 days of an adverse decision, plus a route to human review. The law drops the explicit duty of care but keeps liability under existing anti-discrimination law.
What do US and Canada implications look like for a multi-region retailer?
A national retailer sees four enforcers: the NYC DCWP, the California Civil Rights Department, the Illinois Department of Human Rights and the Colorado Attorney General. Federal anti-discrimination law, enforced by the EEOC, sits underneath all of them, so a state with no AI-specific rule still carries disparate-impact risk when a tool screens out a protected group.
In Canada, Ontario requires employers with 25 or more employees to say in a publicly advertised job posting whether AI is used to screen, assess or select applicants. Osler’s summary of the new job-posting disclosure requirement explains how employers are handling it; verify the current threshold and wording. Quebec’s Law 25 adds a separate duty (verify current text) to inform people when a decision is made exclusively by automated processing. Amounts in the US rules are in US dollars.
On the framework side, the voluntary NIST AI Risk Management Framework (Govern, Map, Measure, Manage) gives HR a common vocabulary for the audit program below. ISO/IEC 42001 plays the same role for a certifiable management system. Fruggr’s AI regulations page tracks how these rules fit together.
How does a CHRO build one audit program instead of five?
Build to the strictest rule in each category, then apply it everywhere. Five steps work for a retail HR team:
- Inventory every tool. Ask HR, store operations and procurement which tools touch candidates. Include tools that staff adopted on their own: shadow AI in recruiting is common.
- Classify each tool by decision. Does it screen, rank, schedule or only communicate? Only the first two are likely in scope everywhere.
- Run a bias audit per tool and per year. Calculate selection rates and impact ratios by sex, race and ethnicity, and keep unknown categories visible.
- Standardize the notice. One plain-language notice, delivered at least 10 business days before first use, meets New York and covers the pre-use notice in Illinois and Colorado.
- Keep records for four years. That satisfies California and exceeds Colorado’s three.
A single register of tools, audits, notices and vendor contracts is what a regulator will ask for. The Trusted AI approach is to keep that register current, not rebuild it each hiring season.
What should a retail HR team do this week?
Name an owner for each tool. Pull the vendor’s most recent bias-audit summary and check its date. Draft the candidate notice once, in plain language. Check job postings in Ontario for the AI disclosure line. Put the Colorado deadline of January 1, 2027 in the calendar and watch the Attorney General’s final rules.

Key takeaways
- One inventory of every hiring tool is the base of compliance in every jurisdiction.
- New York is the strictest on audits and notice; California is the strictest on records; Colorado adds an adverse-decision explanation from January 1, 2027.
- Build one program to the strictest rule in each category instead of five separate ones.
Next step: See how Fruggr maps your hiring tools to NYC Local Law 144 requirements.
FAQ
Software that scores, ranks, screens or recommends candidates or employees, and that people use to make or support a hiring or promotion decision.
Often yes. California extends liability to an employer’s agent, and NYC puts the audit and notice duties on the employer or employment agency using the tool. Check your vendor contracts.
No. New York City requires an annual independent bias audit. California treats anti-bias testing as relevant evidence rather than a mandate, and Colorado requires notice and explanation rather than a published audit. Verify current text.
January 1, 2027, with the Attorney General’s rules due before that date.
Only if they advertise jobs publicly in Ontario and meet the employee threshold. Verify the current wording.
This article was drafted with AI assistance and reviewed by our editorial team before publishing.