When an Algorithm, Not a Person, Makes the Decision Against You

If an automated tool screens job applicants, prices insurance, scores a loan application, or flags a tenant, the underlying anti-discrimination law is generally the same one that would apply to a human decision-maker: Title VII of the Civil Rights Act of 1964 for employment, the Equal Credit Opportunity Act (ECOA) for credit, and the Fair Housing Act for housing, among others. The fact that a vendor's algorithm — not a hiring manager — produced the adverse decision does not exempt the employer or lender from these laws; it can just make the discrimination harder to detect, because a facially neutral scoring model can still produce a discriminatory effect on a protected group (disparate impact).

In May 2023, the EEOC published technical guidance treating algorithmic hiring tools as a Title VII "selection procedure" subject to the same adverse-impact analysis as any other test. That specific guidance document was removed from the EEOC's website in January 2025 following a change in agency leadership. Removing a guidance document does not repeal the underlying statute — Title VII, the ADA, and the ADEA still apply to AI-driven employment decisions — but it does mean there is currently less formal federal agency guidance on how those laws apply specifically to algorithmic tools. verify the EEOC's current published position before relying on any specific AI-hiring standard as official federal guidance

For credit decisions, Regulation B (which implements ECOA) already requires a creditor to give a specific, accurate adverse-action notice when it denies credit — including when a machine-learning model drove the decision. A lender using an opaque algorithm does not get to skip this notice or give a vague reason; the specific factors that actually affected the decision generally still need to be disclosed. verify how your specific lender's adverse-action notice describes the reasons, and whether it appears to actually reflect the model's real decision factors

New York City's Local Law 144 (effective enforcement since July 2023) requires employers using an "automated employment decision tool" on NYC candidates or employees to have the tool independently bias-audited within the past year, publish a summary of the audit results, and give candidates advance notice with the right to request an alternative process. This is currently one of the most concrete, enforceable AI-specific transparency rules in the US, with civil penalties for noncompliance. verify current penalty amounts and enforcement posture, as these can be updated

Colorado passed a broad AI Act (SB 24-205) aimed at algorithmic discrimination in "high-risk" decisions, but its effective date was pushed back twice by the legislature and it was then replaced entirely by a narrower law (SB 26-189, signed May 2026) that takes a disclosure-and-transparency approach instead of the original duty-of-care and impact-assessment framework, effective January 1, 2027. This is a clear example of how fast this area is moving — do not assume the originally-announced Colorado AI Act rules are the ones actually in force. verify Colorado's current AI law and effective date directly before relying on it

In the EU, the AI Act (Regulation (EU) 2024/1689) classifies AI systems used in recruitment, candidate screening, and biometric identification as "high-risk," carrying obligations like risk management and human oversight. Following a 2026 simplification ("omnibus") agreement, those high-risk obligations for most such systems were deferred to 2 December 2027, while narrower transparency rules (like AI-generated content labeling) still take effect in August 2026. verify the current EU AI Act compliance calendar before assuming any specific obligation is already active

Cases where you can show a clear, documented pattern — a bias audit showing a disparity, or a clearly discriminatory rejection reason in an adverse-action notice — have a much stronger footing than a suspicion that "the algorithm seemed unfair," which is hard to prove without underlying data.

NYC's Local Law 144 has driven real transparency (published bias-audit summaries) but oversight assessments have found inconsistent compliance across employers — a missing or low-quality audit is itself worth flagging, but don't assume every employer using an automated tool has actually complied.

This is a genuinely unsettled area of law with rules changing year to year (as the Colorado and EU examples above show) — this page can help you identify the right underlying claim and process, but it cannot predict how a specific agency or court will treat a specific algorithmic tool.