U.S. AI Safety Institute Partners with Anthropic, OpenAI for AI Safety

The U.S. Artificial Intelligence Safety Institute, a key entity within the U.S. Department of Commerce’s National Institute of Standards and Technology (NIST), has entered into significant agreements with leading AI firms Anthropic and OpenAI. These groundbreaking partnerships are set to enhance the development of safe and trustworthy AI technologies.

Framework for Collaborative AI Safety Research

Through Memoranda of Understanding with each company, the U.S. AI Safety Institute will gain early access to new AI models developed by Anthropic and OpenAI, both before and after their public launch. This collaboration will enable focused research on evaluating these models' capabilities, identifying potential safety risks, and developing effective strategies to mitigate those risks.

Advancing the Science of AI Safety

Elizabeth Kelly, director of the U.S. AI Safety Institute, emphasized the importance of these agreements: “Safety is essential to fueling breakthrough technological innovation. With these agreements in place, we look forward to beginning our technical collaborations with Anthropic and OpenAI to advance the science of AI safety. These agreements are just the start, but they are an important milestone as we work to help responsibly steward the future of AI.”

Global Collaboration for AI Safety

The U.S. AI Safety Institute will also work closely with the U.K. AI Safety Institute to provide comprehensive feedback on improving the safety of AI models developed by Anthropic and OpenAI. This international partnership highlights the global importance of ensuring that AI technologies are developed and deployed safely.

NIST’s Legacy in Science and Technology

Building on NIST’s extensive history in advancing measurement science, technology, and standards, the U.S. AI Safety Institute will use these agreements to further its work in AI. The research and evaluations conducted under these partnerships will contribute to a deeper understanding of advanced AI systems and the various risks they present.

Supporting U.S. AI Policy

These agreements align with the Biden-Harris administration’s initiatives on AI, including the recent Executive Order on AI. They also support the voluntary commitments made by leading AI developers to ensure that AI technologies are developed and used in a safe, secure, and trustworthy manner. The collaboration between the U.S. AI Safety Institute, Anthropic, and OpenAI will play a crucial role in advancing these goals.