US Foundation Model Transparency Bill
AI Foundation Model Transparency Act of 2026
United States
RAI-US-NA-HR80940-2026H.R. 8094
US Foundation Model Transparency Bill is Proposed in United States, according to govinfo.gov. We have not yet been able to confirm the status.
BillTransparency and DisclosureGovernance and OversightAccountability and DocumentationH.R. 8094 would establish FTC-led transparency requirements for high-impact AI foundation models.
Summary
H.R. 8094 would direct the FTC, in consultation with NIST, Commerce, and OSTP, to establish transparency requirements for high-impact AI foundation models. Proposed disclosures cover training data, training mechanisms, capabilities, and user-data collection during inference, with protections for small deployers and researchers.
Full article
Read full text ↗Overview
The AI Foundation Model Transparency Act of 2026 is a proposed federal bill identified as H.R. 8094 in the 119th Congress, which covers 2025–2026. Representatives Don Beyer, Mike Lawler, and Sara Jacobs introduced the bipartisan proposal on March 26, 2026. The bill would establish a federal transparency framework addressing how AI foundation models are built, trained, and deployed. Its central mechanism is a direction to the Federal Trade Commission to establish disclosure requirements for high-impact foundation models.
The proposal addresses the information available to consumers about the systems they use and the information available to creators about training data. It would require information to be provided to the FTC and made publicly available. The initiative connects greater visibility into model development and deployment with more informed decisions about AI use. Its proposed disclosure framework focuses on training-data sources, training mechanisms, model capabilities, and collection of user data during inference. As proposed legislation, the bill describes a prospective framework rather than an operative federal disclosure regime.
Foundation Models and Core Concepts
Foundation models are AI models trained on vast and diverse datasets that power many generative AI tools. The congressional announcements identify ChatGPT, Claude, Gemini, and Grok as examples of tools powered by this technology. The breadth of their training and their use across different applications help explain the proposal's focus on information about model development. Training-data sources describe where the information used to train a model originates, while training mechanisms concern the processes through which that information is used to develop the model.
Inference is the use of a model to produce outputs. The proposal treats whether user data is collected during inference as a disclosure subject, distinguishing information collected during model use from information used in training. Model capabilities are another disclosure category, addressing what a model can do. The initiative targets high-impact foundation models and includes a stated objective of protecting small deployers and researchers. These concepts organize the proposed transparency framework around model characteristics, deployment practices, and the participants responsible for making information available.
Governance and Institutional Framework
The FTC would lead the establishment of transparency requirements. Its proposed role encompasses identifying information that high-impact foundation models must provide to the agency and information that must be available to the public. This creates two disclosure audiences: a federal oversight institution and consumers or other members of the public. The distinction is important to the proposal's design because regulator-facing reporting and public-facing transparency serve related but different functions. The agency would establish the requirements rather than leaving the disclosure framework solely to voluntary statements by model deployers.
The FTC would consult the National Institute of Standards and Technology, the Secretary of Commerce, and the Office of Science and Technology Policy. This consultation structure brings the named federal institutions into the development of the transparency framework while keeping the FTC at its center. The Beyer congressional announcement and the Jacobs congressional announcement describe this FTC-led institutional arrangement. The proposal's governance focus is the establishment of disclosure standards and the availability of information about influential AI systems.
Training and Deployment Disclosures
The proposed disclosures include information about training-data sources and a sufficiently detailed summary of the data used to train a foundation model. These categories would give consumers and creators greater visibility into the information underlying model development. A training-data summary serves a different informational purpose from a description of training mechanisms: the former concerns the data used, while the latter concerns how the model is trained. Together, the categories would help explain both the informational inputs and the development processes associated with a foundation model.
The proposal also includes disclosures about model capabilities and whether user data is collected during inference. Capabilities information would help users understand the functions a model offers and assess its suitability for a particular application. Information about inference-time collection would make the treatment of user interactions more visible. The announced framework therefore spans development and deployment, rather than concentrating exclusively on training. Public availability and provision of information to the FTC would support both consumer understanding and institutional oversight of the categories covered by the transparency requirements.
Policy Objectives and Consequential Uses
The proposal responds to concerns about inaccurate, imprecise, or biased AI outputs and about the limited visibility users have into model training and testing. Its proponents identify health-related AI inferences, lending, federal grant approval, housing approval, and law enforcement as examples of consequential applications. In those settings, misunderstanding model capabilities or limitations can affect decisions with significant consequences. The bill's transparency objective is to give users information that helps them judge whether a model is appropriate for an intended use and better understand the basis and limitations of its outputs.
Artists and creators are another group central to the policy rationale. Greater information about training-data sources and the data used in model development would address their interest in understanding what underlies generative AI tools. The proposal also seeks to protect small deployers and researchers while focusing on the highest-impact foundation models. This combination reflects an effort to improve visibility into influential systems while taking account of smaller participants in AI development and deployment. The bipartisan initiative presents transparency as a means of supporting informed use, accountability, and responsible innovation.
Implementation and Organizational Preparation
The proposed implementation mechanism is the FTC's establishment of transparency requirements in consultation with the named federal institutions. Foundation model deployers would make specified information publicly available and provide information to the agency. The framework would turn the bill's transparency objective into operational disclosure requirements if enacted. Its principal preparation themes follow the announced information categories: knowing which foundation models an organization deploys, locating information about their training, documenting capabilities, and understanding data collection during model use.
Voluntary preparation can begin with an inventory of model deployments and the parties that hold relevant documentation. An organization can distinguish records supplied by a model provider from records describing its own deployment practices. It can organize available information on training-data sources and training mechanisms, identify evidence supporting capability descriptions, and map whether inference interactions collect user data. These activities are readiness measures, not duties presently imposed by H.R. 8094. They can help an organization understand the information it possesses and prepare to separate public-facing disclosures from information intended for the FTC.
Monitoring, Evaluation, and Documentation
The proposal's transparency rationale includes helping users assess model suitability and recognize limitations, potential biases, or misleading results. Information about how models are trained and tested is relevant to those judgments, alongside descriptions of capabilities and deployment practices. Making relevant information accessible can support scrutiny of the claims associated with a model and the settings in which it is used. The initiative connects documentation with accountability by seeking greater visibility into development and deployment, especially where inaccurate or biased outputs could influence consequential decisions.
As voluntary preparation, organizations can review whether descriptions of training, capabilities, and inference-data collection are accurate, understandable, and traceable to supporting records. They can compare technical documentation with statements intended for consumers and identify where the two need to be reconciled. Maintaining a clear distinction between provider documentation and deployment-specific information can also improve the reliability of future disclosures. Such reviews support the transparency objective without converting preparatory practices into current bill-imposed requirements. Their practical purpose is to make model information easier to organize, explain, and substantiate.
Relationship to Other Instruments
The bill is a transparency and consumer-protection proposal with an FTC-led institutional structure. Its consultation arrangements include NIST, the Secretary of Commerce, and OSTP, placing several federal institutions within the proposed development of requirements. Its disclosure subjects intersect with broader AI governance concerns, including documentation, model reliability, creators' interests, and user-data practices. Training-data transparency, capability descriptions, and inference-data collection are the particular information categories around which this instrument is organized.
The Beyer announcement also discusses the GUARDRAILS Act, CREATE AI Act, Federal Artificial Intelligence Risk Management Act, and Artificial Intelligence Environmental Impacts Act as part of the representative's broader legislative activity. H.R. 8094's distinctive focus is foundation model transparency through an FTC-led framework. Organizations preparing for that framework can keep its disclosure categories separate from their assessments of other applicable instruments. This preserves the distinction between information about model training and deployment and the wider range of legal and governance issues that can arise from AI development or use.
Domestic Scope and International Policy Context
H.R. 8094 is a United States federal legislative proposal. Its institutional framework centers on domestic federal bodies: the FTC, NIST, the Secretary of Commerce, and OSTP. Its public transparency objective concerns information about foundation models, while its policy rationale includes the role of trustworthy AI in technological leadership. Representative Jacobs links trust in AI to international technological competition. Representative Lawler connects responsible innovation with United States leadership and national security. These themes place the domestic transparency initiative within a broader discussion of technological competitiveness.
The proposal's international policy context complements its consumer and creator focus. Better information about training, capabilities, and model use is presented as relevant to confidence in AI, while responsible innovation is associated with national leadership. For organizations operating in several jurisdictions, clearly organized training and deployment documentation can also be useful when assessing different information requirements separately. Within this instrument, the central mechanism remains the proposed federal transparency framework and the consultation structure assigned to the named United States institutions.
Implementation Timeline
| Milestone | Date or Stage | Description |
|---|---|---|
| Congressional session | 119th Congress, 2025–2026 | H.R. 8094 is the House bill identifier for the AI Foundation Model Transparency Act of 2026. |
| Introduction | 2026-03-26 | Representatives Don Beyer, Mike Lawler, and Sara Jacobs introduced the bipartisan proposal. The official introduction announcement is dated March 26, 2026. |
| Transparency framework development | Proposed implementation stage | The bill would direct the FTC to establish transparency requirements in consultation with NIST, the Secretary of Commerce, and OSTP. |
| Disclosure implementation | Prospective operation | The proposed framework would identify information for public availability and information to be provided to the FTC. |
Preparation Checklist
| Check | Voluntary Preparation |
|---|---|
| Track legislative status | Use the official congressional record to follow H.R. 8094 and distinguish proposed requirements from operative obligations. |
| Review coverage | Assess the legislative text when evaluating covered entities, high-impact models, and treatment of small deployers and researchers. |
| Inventory models | Identify deployed foundation models and the parties holding information about training and deployment. |
| Map training information | Organize available records describing training-data sources, training-data summaries, and training mechanisms. |
| Document capabilities | Collect evidence supporting descriptions of model capabilities and limitations, and compare technical records with user-facing statements. |
| Review inference-data collection | Determine whether user data is collected during model use and record the relevant deployment practices. |
| Separate disclosure audiences | Organize information so that public-facing disclosure and FTC-facing reporting can be assessed separately. |
| Follow implementation | If enacted, align preparation with the applicable FTC requirements developed through the proposed consultation structure. |
Sources and References
| Source | Type |
|---|---|
| H.R. 8094 — AI Foundation Model Transparency Act of 2026 | official |
| H.R. 8094 — Legislative Text | official |
| Beyer, Lawler, Jacobs Introduce Bipartisan Legislation to Promote AI Foundation Model Transparency | official |
| Jacobs, Beyer, Lawler Introduce Bipartisan Legislation to Promote AI Foundation Model Transparency | official |
| Congressional Record Index 2026 — AI Foundation Model Transparency Act | official |
Related Regulations
More AI regulation in United States
AI regulation in United States: full overview
- US AI Security and Innovation Act
- US AI Innovation and Security Executive Order
- NIST AI Risk Management Playbook
- NAIC AI Insurer Guidelines
- United States - National AI Initiative (Pub. L. 116–283)
- United States - AI Research Strategy (2021)
- United States - National AI Research Pilot
- US DOT Automated Vehicles Plan
© Regulations.AI — created on 21 Nov 2026 using Gemini 3.6 Flash