Algorithmic Accountability Act of 2025

Bill Number

S. 2164

Origin Chamber

Senate

Congress

119th Congress, Session 1

Policy Area

Commerce

Status

Introduced

Latest Action

2025-06-25: Read twice and referred to the Committee on Commerce, Science, and Transportation.

Last Updated

2025-07-18T14:51:14Z

AI-Generated Summary

Algorithmic Accountability Act of 2025 (S. 2164)

Purpose

The legislation aims to promote accountability in the use of automated decision systems (ADS)—software or processes using AI, machine learning, or data techniques to inform decisions—and augmented critical decision processes (ACDP), which apply ADS to make important choices affecting consumers' lives. It directs the Federal Trade Commission (FTC) to require impact assessments to identify and mitigate risks like bias, privacy harms, and unfair outcomes in areas such as employment, healthcare, housing, and financial services.

Key Provisions

• Definitions and Scope:
• Defines ADS as computational tools (excluding basic infrastructure like data storage) that influence decisions.
• Defines ACDP as processes using ADS for "critical decisions," which have significant effects on access to or costs of essentials like education, jobs, utilities, healthcare, housing, or financial services.
• "Covered entities" include businesses under FTC jurisdiction that are large (e.g., over $50 million in annual gross receipts or handling data on 1 million+ consumers) or meet related thresholds; smaller developers of ADS for large users are also included, with thresholds adjusted for inflation.

• Regulatory Requirements:
• Within 2 years of enactment, the FTC must issue rules (after consulting experts like NIST and civil rights advocates) mandating covered entities to conduct ongoing impact assessments for ADS and ACDP before and after deployment.
• Assessments must evaluate risks (e.g., bias based on race, gender, or age; privacy protections; performance across groups), consult stakeholders (e.g., employees, advocates), document data sources and decisions, and mitigate negative impacts.
• Covered entities must maintain assessment records for 3 years post-deployment and submit annual summary reports (plus initial reports for new systems) to the FTC in a standardized, machine-readable format.
• Non-covered entities can voluntarily submit reports.
• Entities must disclose their covered status to vendors and attempt to fix harmful impacts.

• Reporting and Transparency:
• FTC must create a public online repository of anonymized, limited summary data (e.g., company names, decision types, opt-out links) updated quarterly to inform consumers and researchers, while protecting business secrets.
• FTC must issue annual public reports on trends from submissions to guide oversight and recommendations.

• Support and Enforcement:
• FTC to provide guidance, templates, and technical assistance; establish a new Bureau of Technology (with 50+ staff) and add 25 enforcement personnel.
• Violations treated as unfair/deceptive practices under FTC Act; enforced by FTC (civil penalties) and states (with notice to FTC); no private right of action, but states can sue on behalf of residents.
• Coordinates with other agencies; no preemption of state/local laws; periodic review of rules every 5 years.

Significant Changes to Existing Law

• Introduces first federal mandate for pre- and post-deployment impact assessments of AI-driven decisions, building on FTC's existing authority over unfair practices (15 U.S.C. § 45) but adding specific AI-focused requirements like bias testing and stakeholder consultation.
• Creates new FTC infrastructure (Bureau of Technology) and reporting mechanisms, absent in prior laws like the FTC Act.
• Expands state enforcement roles while preserving FTC primacy; explicitly allows voluntary reporting and public disclosure beyond requirements, differing from narrower privacy laws like HIPAA or GDPR equivalents.

Potential Impacts

• Government Agencies: Increases FTC workload with rulemaking, reviews, and a new bureau, requiring appropriations; fosters inter-agency coordination (e.g., with NIST, OSTP) for AI standards, potentially streamlining oversight but raising administrative costs.
• Citizens (Consumers): Enhances protections against discriminatory or opaque AI decisions in daily life (e.g., job screenings, loan approvals), providing better notice, opt-outs, and recourse; public repository aids awareness and research, but full assessments remain private to avoid privacy risks.
• Businesses: Imposes compliance burdens (assessments, reporting) on large tech/finance firms, possibly increasing costs for AI development; encourages ethical practices and innovation in bias mitigation, with flexibility for infeasible requirements.
• International Relations: Minimal direct impact, but could influence global AI norms by setting U.S. standards for accountability, potentially affecting trade in AI tech or harmonizing with EU AI Act; no extraterritorial provisions.

Main Stakeholders Affected

• Covered Entities: Large corporations (e.g., tech giants like Google, financial firms like banks) deploying AI for critical decisions; smaller AI developers selling to them.
• Consumers: Individuals impacted by AI in employment, healthcare, housing, etc., gaining indirect protections.
• Advocates and Researchers: Civil rights groups, academics, and tech experts benefiting from consultations, public data, and transparency.
• Government: FTC (primary enforcer), states (litigation support), and agencies like NIST (consultation on standards).
• Third Parties: Vendors, partners, and recipients of AI outputs (e.g., employers using hiring software).

Notable Legal, Constitutional, or Political Implications

• Legal: Treats violations as unfair practices, enabling FTC fines up to $50,120 per violation (adjusted); emphasizes mitigation without mandating bans, balancing innovation with harm prevention; rule of construction allows extra voluntary measures.
• Constitutional: May raise due process concerns if assessments limit speech (e.g., AI explanations), but focuses on internal documentation; no direct First Amendment conflict as it targets commercial practices, not content.
• Political: Bipartisan sponsors highlight growing AI regulation consensus; empowers FTC without new funding mandates beyond authorizations, potentially sparking debates on overreach vs. consumer safety; non-preemptive stance preserves state innovation (e.g., California's AI laws).

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

Sponsor

Sen. Wyden, Ron [D-OR]

Cosponsors (7)

Sen. Warren, Elizabeth [D-MA], Sen. Booker, Cory A. [D-NJ], Sen. Heinrich, Martin [D-NM], Sen. Luján, Ben Ray [D-NM], Sen. Merkley, Jeff [D-OR], Sen. Hirono, Mazie K. [D-HI], Sen. Schatz, Brian [D-HI]

Recent Actions

• 2025-06-25: Read twice and referred to the Committee on Commerce, Science, and Transportation.
• 2025-06-25: Introduced in Senate

Bill Versions

• Algorithmic Accountability Act of 2025 — issued 2025-06-25 — PDF (50 pages)