Sponsors (4)
The bill increases transparency and strengthens safety/national‑security oversight of powerful AI models, but does so by imposing compliance costs and disclosure risks that may favor large, well‑resourced firms and lead to partial withholding of information.
Consumers (including students and families) and downstream deployers gain clearer, public disclosures about model capabilities, limitations, and supported languages, making it easier to understand risks and use models safely.
Public safety and national security oversight is strengthened because covered entities must disclose precautions for high‑risk categories (e.g., bio, medical, national security), improving regulators' and first responders' situational awareness.
Developers, researchers, and auditors receive standardized, machine‑readable model metadata, facilitating comparison, auditing, and safer deployment of models across organizations.
Model providers (especially smaller ones) face increased compliance costs and administrative burdens to collect, document, and publish extensive training, governance, and performance data.
Compliance complexity and reporting burdens may advantage well‑resourced firms and disadvantage smaller developers despite assistance and grace periods, concentrating market power.
Public disclosure requirements could create intellectual property and competitive risks, leading firms to withhold detail or reduce innovation to protect trade secrets.
Based on analysis of 2 sections of legislative text.
Mandates FTC rules (within 1 year) requiring foundation-model providers to submit detailed transparency disclosures and post them publicly in human- and machine-readable formats.
Introduced March 26, 2026 by Donald Sternoff Beyer · Last progress March 26, 2026
Requires the Federal Trade Commission to write rules (within 1 year) that force providers of foundation AI models to disclose detailed information about each model to the FTC and to the public in both human- and machine-readable formats. The rules must be developed with input from NIST, Commerce, OSTP and other stakeholders, allow sensitive material to be submitted privately, and permit disclosures via model cards or similar documents. The disclosure list includes training-data summaries, whether data are collected/retained during inference, broad data composition descriptions (demographics, languages), data-governance practices, intended uses and limitations, version history and knowledge cutoff, incident monitoring and response, supported languages, alignment efforts, and evaluation results — plus processes and guidance the FTC will provide to help entities comply.