OpenAI and Anthropic have agreed to provide undisclosed models to the US AI Safety Institute.

August 30, 2024 (excerpt)

OpenAI and Anthropic have agreed to provide undisclosed artificial intelligence (AI) models to the US government's AI Safety Institute. This groundbreaking move is a crucial step in promoting the safety and responsible development of AI. OpenAI and Anthropic have established a formal collaboration with the US AI Safety Institute, which is located within the National Institute of Standards and Technology (NIST) of the US Department of Commerce. Under this agreement, both companies will provide the institute with new AI models for evaluation, both before and after their public release, to undergo safety assessments.
This initiative is part of the activities of the US AI Safety Institute, which was established in 2023 following an executive order issued by President Biden. The institute aims to collaborate with companies and experts to establish safety guidelines and best practices for AI systems.