What LL 144 requires
NYC Local Law 144 regulates the use of Automated Employment Decision Tools (AEDTs) — systems that substantially assist or replace discretionary decision-making in employment.
The law has three core obligations:
- An independent bias audit, conducted within the prior year by an independent auditor (not the employer, not the vendor).
- Public summary disclosure of the audit results on the employer's website.
- Candidate notice at least 10 business days before the AEDT is used, describing the job qualifications and characteristics the tool will evaluate.
The DCWP-published rules (5 RCNY §5-300 et seq.) specify the audit methodology in detail: impact-ratio analysis across race / ethnicity, sex, and intersectional combinations, against the four-fifths rule.
Who's affected
LL 144 applies to:
- NYC employers using an AEDT, OR
- Out-of-NYC employers using an AEDT for candidates physically in NYC or for NYC-based roles.
The "physically in NYC" branch matters: a fully remote role advertised to candidates anywhere triggers LL 144 if a candidate is sitting in NYC when they apply.
Key deadlines
| Date | Trigger |
|---|---|
| 5 July 2023 | Enforcement began. |
| Rolling | Bias audits are valid for one year. A current audit must be on file at all times of AEDT use. |
LL 144 has been in force the longest of any US AI hiring regime. Its audit methodology has effectively become the US benchmark — Colorado's impact assessment and Illinois HB 3773 disparate-effect testing both lean on the same statistical apparatus.
What hiring deployers specifically must do
- Identify whether your tool is an AEDT. The DCWP rules give a narrower definition than the original statute: the tool must produce a "simplified output" (score, classification, recommendation) that is used as a "substantial factor". A pure search/sort interface is generally not an AEDT.
- Commission an independent bias audit within the prior year. "Independent" means the auditor has no financial relationship with the employer or the vendor. Major auditors as of 2026 include BABL AI, DCI Consulting, ORCAA, ConductorAI, Holistic AI, Warden AI, Credo.AI, Secretariat, and Parity.
- Publish a summary on your website including the date of the most recent audit, the data source (employer or historical, or both), and the impact-ratio results across required subgroups.
- Provide candidate notice 10+ business days in advance, listing the job qualifications evaluated and offering an alternative process on request.
- Keep records for DCWP enforcement requests.
Common misconceptions
- "The vendor's audit is enough." Sometimes. If the vendor's audit used data from a population similar to the deployer's and the deployer is using the tool as audited, the deployer can rely on the vendor audit. Many employers commission their own anyway for defensibility.
- "The audit lives in the vendor's trust portal." Frequently — and that's a problem. LL 144 requires the summary be on the employer's website. A trust portal link is not equivalent.
- "NYC LL 144 covers all AI in hiring." No. AEDT is narrowly defined. A conversational scheduler that does no decisioning is not in scope; a resume-ranker that produces a "fit score" is.
- "Once published, the audit is good forever." The audit must be within the prior year. Audits expire annually.
How vendors are addressing it
LL 144 is the regulation where vendor posture is most visible. Across the directory:
- HireVue publishes annual DCI Consulting audits since 2023 (plus an ORCAA audit in 2021).
- Eightfold, Harver, and Pymetrics (Harver Games) publish BABL AI audits with intersectional analyses; Harver runs separate audits for the Harver Platform and the pymetrics product.
- Phenom publishes ConductorAI audits annually.
- Beamery uses Warden AI for continuous monthly bias monitoring with a publicly viewable dashboard — the only vendor in the peer set with a continuous (not annual point-in-time) audit posture.
- Workday / HiredScore publishes a Secretariat audit whose conclusions are openly disputed by plaintiffs in Mobley v. Workday.
- Paradox and SeekOut rely on vendor audits that are not publicly downloadable; deployers using these tools should commission their own.
See Bias Audit Transparency in the rubric and the per-vendor profiles for cited links.