The AI Foundation Model Transparency Act of 2026 mandates the Federal Trade Commission (FTC) to establish comprehensive regulations for artificial intelligence foundation models. These regulations aim to significantly improve transparency regarding the training data, documentation, testing, and operations of these models, both before commercial deployment and throughout their lifecycle. The bill defines "covered entities" as providers of foundation models that meet specific criteria, such as exhibiting high-risk potential, having over 10 million monthly users, or being trained with immense computational power. Under the proposed regulations, covered entities will be required to submit detailed information about each foundation model to the FTC and also make certain information publicly available . This public disclosure must be in both human-readable and machine-readable formats, with some data hosted on the entity's website and other key information centrally located on an FTC-hosted website. The bill specifies a broad range of information to be disclosed, including summaries of training data sources, data governance procedures, intended purposes and limitations, and performance under various benchmarks. Key disclosures include descriptions of the size and composition of training data, efforts to align with the NIST AI Risk Management Framework , and performance evaluations, particularly for models addressing high-risk areas like medical questions, national security, or elections. The bill also allows for redactions to protect cybersecurity, public safety, national security, or to comply with federal law. While fully open-source models are exempt, the FTC is directed to provide assistance and grace periods for small and new businesses to facilitate compliance with these new transparency requirements.