Artificial Intelligence Civil Rights Act of 2025

Bill Number: S. 3308
Origin Chamber: Senate
Congress: 119th Congress, Session 1
Status: Introduced
Latest Action: 2025-12-02: Read twice and referred to the Committee on Commerce, Science, and Transportation.
Last Updated: 2025-12-24T03:53:20Z

AI-Generated Summary

Artificial Intelligence Civil Rights Act of 2025 (S. 3308)

Purpose

This legislation aims to protect individuals' civil rights by preventing discrimination, harm, and unfair practices caused by artificial intelligence (AI) algorithms in important life decisions, known as "consequential actions." It ensures fairness, transparency, and accountability in AI development and use, building on existing civil rights laws to address biases in automated systems.

Key Provisions

The Act defines key terms like "covered algorithm" (AI or machine learning processes used in consequential actions such as hiring, lending, or policing), "consequential action" (decisions affecting access to jobs, housing, healthcare, etc.), "protected characteristic" (traits like race, sex, disability, or age), and roles like "developer" (AI creators) and "deployer" (AI users in business).

Title I: Civil Rights
Prohibits developers and deployers from using covered algorithms in ways that discriminate or create unjustified unequal impacts based on protected characteristics, denying equal access to opportunities.
Requires pre-deployment evaluations (before AI use) to assess risks of harm or bias, conducted by independent auditors if harm is plausible; includes testing data, training methods, and stakeholder consultations.
Mandates annual post-deployment impact assessments for deployers to monitor real-world effects, with developers reviewing these summaries yearly.
Exceptions allow self-testing for compliance or non-commercial research; private clubs are exempt.

Title II: Covered Algorithm and Contract Standards
Developers and deployers must mitigate identified risks, certify AI safety and benefits outweigh harms, and ensure reliable performance.
Bans deceptive marketing of AI and "off-label" use (using AI beyond tested purposes without assuming developer duties).
Contracts between developers and deployers must detail data handling, prohibit unauthorized data mixing, and allow concern-raising without retaliation.
Provides rights to human alternatives (opt-out of AI for key decisions), appeals to humans, and protections against retaliation for exercising rights or whistleblowing.
FTC to issue rules within 2 years on opt-outs and appeals, considering accessibility and feasibility.

Title III: Transparency
Requires public disclosures on AI practices, including data use, transfers, rights exercises, and evaluation summaries, in plain language and covered languages (top 10 U.S. languages).
Short-form notices for users upon AI interaction; mechanisms for reporting violations.
FTC to study explanations of AI decisions (e.g., why a loan was denied) and report to Congress within 18 months.
FTC to create a consumer rights webpage, annual reports on assessments, and a public repository for evaluations (with privacy redactions).

Title IV: Enforcement
Treats violations as unfair/deceptive practices under FTC Act; FTC enforces with expanded jurisdiction over nonprofits, banks, etc.
States' attorneys general can sue for injunctions, penalties (up to $15,000 per violation or 4% of revenue), and damages.
Private right of action allows individuals to sue for treble damages ($15,000 minimum per violation), fees, and relief; invalidates pre-dispute arbitration waivers for these claims.
Notice to FTC/states required before private suits; severability clause ensures partial invalidity doesn't void the whole Act.

Title V: Federal Resources
Creates new federal job series for AI auditing.
Authorizes funds and up to 500 new FTC staff for enforcement.

Significant Changes to Existing Law

Expands civil rights protections (e.g., from Civil Rights Act of 1964) to explicitly cover algorithmic discrimination, including disparate impact (unequal effects without justification) presuming AI as a single unit unless proven separable.
Introduces mandatory AI risk assessments and audits, absent in prior laws, with FTC rulemaking authority.
Grants private lawsuits and state enforcement for AI harms, stronger than many current tech regulations; overrides arbitration for disputes.
Does not lessen other laws (e.g., labor, health) but adds AI-specific duties without conflicting.

Potential Impacts

Government Agencies: FTC gains broad enforcement tools and resources, potentially increasing oversight of AI in sectors like finance and healthcare; states get new litigation powers, fostering coordinated federal-state actions. Other agencies (e.g., for elections, immigration) must comply if using AI.
Citizens: Enhances protections against biased AI decisions in daily life (e.g., job applications, loans), with rights to human review, appeals, and explanations; promotes awareness via notices and repositories, empowering marginalized groups.
International Relations: Focuses on U.S. persons but could set global standards for ethical AI, influencing trade/tech policies; no direct international provisions, but data transfers may affect cross-border AI use.

Main Stakeholders Affected

Developers and Deployers: Tech companies, AI firms, and businesses using AI (e.g., in hiring software, credit scoring) face compliance costs for audits, disclosures, and mitigations; shared responsibilities in contracts.
Individuals: U.S. natural persons impacted by AI decisions, especially those with protected characteristics, gain enforcement tools and rights.
Independent Auditors: New role for objective experts in evaluating AI risks.
Government Entities: FTC, state attorneys general, and agencies using AI (e.g., for benefits, policing) must follow standards; whistleblowers protected.
Communities and Nonprofits: Stakeholder consultations required; civil society may use repository for advocacy.

Notable Legal, Constitutional, or Political Implications

Legal: Strengthens enforcement via private actions and FTC powers, potentially leading to more litigation; defines "harm" broadly (e.g., constitutional rights infringement), aiding disparate impact claims without proving intent.
Constitutional: Aligns with Equal Protection Clause by prohibiting bias in government/private AI use; protects free speech/whistleblowing but doesn't limit labor bargaining rights.
Political: Advances civil rights in the AI era, sponsored by progressive senators; may spark debates on innovation vs. regulation, with industry concerns over costs and trade secrets, while supporting equity for underserved groups. Rules of construction preserve other laws, avoiding preemption challenges.

This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.

Sponsor

Sen. Markey, Edward J. [D-MA]

Cosponsors (4)

Sen. Booker, Cory A. [D-NJ], Sen. Merkley, Jeff [D-OR], Sen. Warren, Elizabeth [D-MA], Sen. Hirono, Mazie K. [D-HI]

Recent Actions

2025-12-02: Read twice and referred to the Committee on Commerce, Science, and Transportation.
2025-12-02: Introduced in Senate

Bill Versions

Artificial Intelligence Civil Rights Act of 2025 — issued 2025-12-02 — PDF (71 pages)