01
What Local Law 144 is trying to prevent
The law targets opaque automated tools that influence who advances without transparency or outcome monitoring. Covered use typically involves tools that substantially assist or replace discretionary decision-making for employment decisions in NYC. If your AI assessment silently ranks or filters candidates, treat compliance as a product requirement-not a legal afterthought.
- Map every AI-assisted step that can advance, hold, or reject a candidate
- Identify whether the tool substantially assists a covered employment decision
- Document vendors, versions, and who owns the hiring decision
- Assume “AI helped write the score” still needs human accountability
02
Bias audits are an operating cadence
A one-time PDF is not a program. Bias audits for AEDTs generally require independent analysis of selection rates and impact ratios across protected categories where data is available, plus a published summary. Build a calendar: freeze assessment versions, export outcome data, run the audit, publish required summaries, and remediate when impact ratios look problematic.
- Freeze question sets and rubrics for each hiring wave you will audit
- Export pass/advance rates by role family and assessment version
- Track reviewer overrides-disagreement is a validation signal
- Publish required summaries on the cadence your counsel advises
03
Candidate notice and transparency
Candidates generally need clear notice when an AEDT is used, plus information about the job qualifications and characteristics the tool measures. Vague “we use AI” banners are not enough. Pair notice with accommodation paths and a human review path for disputed outcomes.
- State what is scored (job-related competencies and evidence)
- State what is not scored (faces, accents, appearance, inherent personality)
- Explain how humans review scores and integrity flags
- Provide a contact path for accommodations and questions