AI Compliance Glossary
Key terms for NYC LL144 & EU AI Act — in plain English
AEDT (Automated Employment Decision Tool)
Any computational process using machine learning, AI, statistical modeling, or data analytics that generates predictions, scores, or classifications about job candidates and is used to substantially assist in hiring or promotion decisions. Defined in NYC Admin Code § 20-870(1).
Adverse Impact
A substantially different rate of selection in hiring, promotion, or other employment decision that works to the disadvantage of members of a protected group. Measured using the four-fifths rule.
Bias Audit
An independent statistical evaluation of an AEDT that measures disparate impact across protected groups (sex, race/ethnicity, intersectional categories). Required annually under NYC LL144.
Candidate Notice
Written notification to candidates that an AEDT will be used to evaluate them, required at least 10 business days before the AEDT is used. Must include job categories and alternative process instructions.
CE Marking
A certification mark indicating conformity with health, safety, and environmental protection standards for products sold within the European Economic Area. Required for high-risk AI systems under the EU AI Act.
Conformity Assessment
The process of verifying that an AI system meets the requirements set out in the EU AI Act. For high-risk AI, this may involve third-party assessment or self-assessment depending on the system type.
DCWP
NYC Department of Consumer and Worker Protection. The agency responsible for enforcing NYC Local Law 144, including investigating violations and assessing penalties.
Disparate Impact
A facially neutral policy or practice that disproportionately affects members of a protected group. In AI hiring, this occurs when an AEDT selects candidates at different rates by race, sex, or other protected characteristics.
Four-Fifths Rule
EEOC guideline stating that a selection rate for any protected group that is less than 80% (0.80) of the rate for the highest-selected group indicates potential adverse impact. Codified at 29 C.F.R. § 1607.
High-Risk AI
Under the EU AI Act, AI systems that pose significant risks to health, safety, or fundamental rights. Includes AI used in employment, hiring, credit scoring, education, healthcare, and biometric identification. Listed in Annex III.
Impact Ratio
The selection rate of a protected group divided by the selection rate of the highest-selected group. An impact ratio below 0.80 indicates potential adverse impact under the four-fifths rule.
Independent Auditor
A person or entity that conducts a bias audit and is not involved in the development or training of the AEDT, not affiliated with the employer or vendor, and has no financial interest in the audit outcome.
Intersectional Categories
Protected groups defined by the intersection of two or more characteristics, such as Black women or Hispanic men. NYC LL144 requires bias audits to report impact ratios for intersectional categories, not just single categories.
Public Disclosure
The requirement under NYC LL144 to publish a summary of the most recent bias audit on a publicly accessible web page, including auditor name, date, selection rates, and impact ratios.
Prohibited AI
AI practices banned outright under the EU AI Act, including social scoring, emotion recognition in workplaces, manipulative AI, and biometric categorization by sensitive attributes. In effect since February 2, 2025.
Risk Management System
A documented process for identifying, assessing, and mitigating risks throughout an AI system's lifecycle. Required for high-risk AI under EU AI Act Article 8.
Selection Rate
The percentage of candidates from a given group who are selected, recommended, or advanced by an AEDT. Used as the basis for calculating impact ratios.
Substantially Assist
A legal standard under NYC LL144 indicating that an AEDT plays a meaningful role in hiring or promotion decisions. Tools that only perform basic keyword search or scheduling are excluded.
Transparency Obligations
Requirements under EU AI Act Article 50 that deployers inform individuals when they are interacting with AI systems. Takes effect August 2, 2026.
Want to Learn More?
Our guides explain these terms in context with examples, scenarios, and source citations.
Browse All Guides