AI Foundation Model Transparency Act of 2026
- Bill Number
- H.R. 8094
- Origin Chamber
- House
- Congress
- 119th Congress, Session 2
- Policy Area
- Commerce
- Status
- Introduced
- Latest Action
- 2026-03-26: Referred to the House Committee on Energy and Commerce.
- Last Updated
- 2026-04-09T17:22:19Z
AI-Generated Summary
Purpose
The AI Foundation Model Transparency Act of 2026 directs the Federal Trade Commission (FTC) to create rules requiring large AI developers to disclose details about the training data, operations, and risks of their "foundation models" (large, general-purpose AI systems trained on vast datasets, like those powering advanced chatbots or image generators). The goal is to increase public transparency before and during commercial use, helping users understand potential biases, risks, and limitations.
Key Provisions
- Rulemaking Timeline: FTC must issue regulations within 1 year of enactment, in consultation with NIST, Commerce Department, OSTP, and stakeholders (e.g., standards groups, AI companies, academics, civil rights advocates). Rules apply 90 days after issuance, with annual updates.
- Required Disclosures for foundation models from covered entities (see Stakeholders):
- Summary of training data sources, collection methods, size/composition (e.g., demographics, languages), and governance.
- Intended uses, risks/limitations, model version, knowledge cutoff date, incident monitoring.
- Supported languages, alignment efforts (e.g., with NIST's AI Risk Management Framework), benchmark performance on high-risk tasks (e.g., health advice, cybersecurity, elections, hiring).
- Compute power used for training/operation.
- Public Availability:
- Human-readable summaries on company websites.
- Machine-readable data on an FTC-hosted central site.
- Some sensitive info (e.g., personal data, cybersecurity details) submitted privately to FTC.
- Option to comply via "model cards" or "system cards" (standardized AI documentation formats).
- Exemptions and Flexibilities:
- Fully open-source models are exempt.
- Downstream (derived) models must link to base model's disclosures and report significant changes.
- Small/new businesses get compliance guidance, templates, a 3-month penalty grace period, and FTC support reps.
- Redactions allowed for security, safety, or legal reasons, with justifications.
- FTC Duties: Create a resources webpage for non-covered AI developers; submit a report to Congress within 1 year.
- Enforcement: Violations treated as unfair/deceptive practices under the FTC Act; 14-day notice before action.
Significant Changes to Existing Law
- Introduces new FTC regulations specifically for AI foundation models, with no prior equivalent federal transparency mandate.
- Expands FTC enforcement powers to certain non-profits and common carriers (e.g., telecoms) for these AI rules, overriding some jurisdictional limits.
- Ties compliance to existing FTC Act penalties, without creating new ones.
Potential Impacts
- Government Agencies: FTC gains rulemaking/enforcement workload, including a public website and small business support; coordinates with NIST/Commerce/OSTP.
- Citizens/Consumers: Greater access to AI risk info (e.g., biases in hiring/medical advice), potentially building trust and enabling better-informed use; protects vulnerable groups (minors, seniors).
- AI Companies: Large developers face disclosure costs/burdens; small ones get relief. Could standardize industry practices via benchmarks and frameworks.
- International Relations: Minimal direct impact, though U.S. standards may influence global AI norms.
Main Stakeholders Affected
- Covered Entities: AI providers under FTC jurisdiction whose foundation models meet thresholds (high-risk capabilities, >10M monthly users/downloads, or >10²⁶ floating-point operations in training).
- Small/New Businesses: AI startups get tailored assistance to ease compliance.
- Consumers and Civil Rights Advocates: Benefit from public disclosures on risks/biases.
- FTC and Federal Partners (NIST, Commerce, OSTP): Lead implementation.
- AI Users/Deployers: Access info on model limitations, especially in high-risk areas like elections or health.
Notable Legal, Constitutional, or Political Implications
- Legal: Relies on standard Administrative Procedure Act (APA) rulemaking (notice-and-comment process); FTC enforces via existing consumer protection authority, avoiding new penalties.
- Constitutional: Potential scrutiny over compelled speech (disclosures), but framed as anti-deception measure; redactions protect security/national interests.
- Political: Bipartisan (introduced by Reps. Beyer, Lawler, Jacobs); promotes U.S. AI leadership through transparency without heavy-handed bans, balancing innovation and safety. Annual reviews allow adaptation to tech advances.
This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.
Sponsor
Rep. Beyer, Donald S. [D-VA-8]
Cosponsors (3)
Rep. Lawler, Michael [R-NY-17], Rep. Jacobs, Sara [D-CA-51], Rep. Fitzpatrick, Brian K. [R-PA-1]
Recent Actions
- 2026-03-26: Referred to the House Committee on Energy and Commerce.
- 2026-03-26: Introduced in House
- 2026-03-26: Introduced in House
Bill Versions
- AI Foundation Model Transparency Act of 2026 — issued 2026-03-26 — PDF (16 pages)