NYC Local Law 144 vs California FEHA AI Rules

No, and the two barely intersect. Local Law 144 is satisfied by obtaining an independent bias audit within the preceding year, publishing a summary of its results and the tool’s distribution date, and giving candidates 10 business days’ notice. California’s Employment Regulations Regarding Automated-Decision Systems require none of those things and are not satisfied by any of them. A California claim asks whether an employment benefit was granted or denied in a way that violates FEHA; an audit is potentially useful evidence toward a defence, but under 2 CCR § 11009(f) its quality, recency, scope, efficacy, results and the employer’s response to those results are all in play. An audit showing an adverse impact that the employer then ignored is worse than no audit at all in California, while in New York it is full compliance.

The rules at 6 RCNY § 5-300 build on the Administrative Code definition and narrow the phrase "to substantially assist or replace discretionary decision making" to three situations: relying solely on a simplified output such as a score, tag, classification or ranking with no other factors considered; using a simplified output as one of a set of criteria where it is weighted more heavily than any other criterion; or using a simplified output to overrule conclusions derived from other factors, including human decision-making. A tool that produces a score a human recruiter treats as one input among several, weighted no more than the others, falls outside that test.

They can. The Civil Rights Council added a definition of "agent" to 2 CCR § 11008 covering any person acting on an employer’s behalf, directly or indirectly, to exercise a function traditionally exercised by the employer — applicant recruitment, screening, hiring, promotion, or decisions about pay, benefits or leave — and it states that an agent "may include, in appropriate circumstances, a third party that creates an automated-decision system used by or for an employer, or a third party that uses an automated-decision system on behalf of an employer." An agent of an employer is also an "employer" for purposes of the Act. The Council considered and declined requests to carve consumer reporting agencies out of that definition, noting they may still meet it to the extent they perform FEHA-regulated activity. The word "may" and the phrase "in appropriate circumstances" are doing real work here — this is a fact-specific test, not an automatic vendor-liability rule.

Not as a rule. A Federal Register full-text search for "artificial intelligence" restricted to documents issued by the Equal Employment Opportunity Commission returns exactly one result: the agency’s Draft Strategic Enforcement Plan, a notice published on 10 January 2023 seeking comment on its 2023–2027 enforcement priorities, which names employers’ increasing use of automated systems to target job advertisements and inform hiring decisions. That is a statement of where the agency intends to look, not a regulation. Existing federal law — Title VII, the ADA, the ADEA, and the 1978 Uniform Guidelines on Employee Selection Procedures — applies to automated selection procedures the same way it applies to any other, which is precisely the theory California codified and New York did not.

Yes, on both coasts, and neither instrument exempts them. Local Law 144 binds employers and employment agencies using an AEDT in New York City, which includes universities and academic medical centres hiring research staff, postdocs or administrative personnel through a screening tool that meets the AEDT test. In California the point is on the record: the Civil Rights Council expressly declined a request from the Alliance for Digital Innovation to exclude public bodies from the rulemaking, reasoning that FEHA’s own definition of employer includes "the state or any political or civil subdivision of the state, and cities." Public university systems are therefore covered as FEHA employers, and the "agent" definition can reach the applicant-tracking or screening vendor an HR or sponsored-programs office procures.

NIKOLAI is CASRAI’s own independent frontier-AI-safety dictionary. It is not endorsed by any organization whose published terms it reads, and every crosswalk row in it is a shadow mapping unless that organization has filed a Mapping Declaration. Its N8 element "Evaluator Independence and Conflict of Interest" records the financial, organisational and personal relationships between an evaluator and the party being assessed, along with the verification process used to clear the evaluator. Local Law 144’s "independent auditor" definition at 6 RCNY § 5-300 is a working legal instance of that same three-part structure: no involvement in using, developing or distributing the tool; no employment relationship with the employer or vendor during the audit; no direct or material indirect financial interest in either. NIKOLAI’s crosswalk for that element does not currently carry a DCWP row, and this page does not claim one exists — the parallel is CASRAI’s structural reading, not New York’s.