Prohibits algorithms used to determine access to important opportunities (e.g., employment) or information about them in Washington D.C. from discriminating on the basis of protected classes; also mandates related algorithm audits, disclosures, and remedies for decisions made based on inaccurate data.
Analysis summaries, actor details, and coverage mappings were LLM-classified and may contain errors.
This is a binding legislative bill with mandatory obligations, civil penalties up to $10,000 per violation, enforcement by the Attorney General, and private right of action. The document uses mandatory language throughout ('shall') and establishes formal enforcement mechanisms.
The document has good coverage of approximately 6-8 subdomains, with strong focus on unfair discrimination (1.1), unequal performance across groups (1.3), privacy compromise (2.1), security vulnerabilities (2.2), lack of transparency (7.4), and lack of robustness (7.3). Coverage is concentrated in discrimination/toxicity, privacy/security, and AI system safety domains.
The document governs algorithmic decision-making across multiple sectors where 'important life opportunities' are determined, with explicit coverage of employment, housing, credit/finance, insurance, education, and public accommodations. The regulation applies to any covered entity making algorithmic decisions in these areas within the District of Columbia.
The document primarily addresses the Deploy and Operate and Monitor stages of the AI lifecycle, with some coverage of Verify and Validate. It focuses on algorithmic systems already in use for decision-making, requiring ongoing audits, impact assessments, and monitoring for discriminatory effects. There is minimal coverage of earlier stages like planning, data collection, or model building.
The document explicitly covers AI systems and algorithmic processes that utilize machine learning and artificial intelligence for decision-making. It does not distinguish between different types of AI (frontier, general purpose, task-specific, foundation models, generative, or predictive) and does not mention compute thresholds or open-weight models. The focus is on algorithmic decision-making systems broadly defined.
Council of the District of Columbia
The bill is proposed by the Council of the District of Columbia, which is the legislative body for the District. The document explicitly states this in the title and findings section.
Attorney General for the District of Columbia; Office of the Attorney General; Superior Court of the District of Columbia
The Attorney General for the District of Columbia is designated as the primary enforcement authority with power to bring civil actions, issue subpoenas, and impose penalties. The Superior Court of the District of Columbia serves as the judicial enforcement body.
Office of the Attorney General for the District of Columbia
The Office of the Attorney General receives annual audit reports from covered entities. Additionally, covered entities themselves are required to conduct ongoing monitoring through annual audits and impact assessments of their algorithmic systems.
The bill targets 'covered entities' which are defined as organizations that make algorithmic eligibility or information availability determinations, or rely on such determinations from service providers. This includes both developers of algorithmic systems and deployers who use them for decision-making about employment, housing, credit, insurance, and other important life opportunities.
9 subdomains (5 Good, 4 Minimal)