Comment on Frontiers in AI for Science, Security, and Technology (FASST) Initiative

Response to Department of Energy

On the 12th of September, the Department of Energy’s (DOE) Office of Critical and Emerging Technologies released a request for information on Frontiers in AI for Science, Security, and Technology (FASST) Initiative. The FASST initiative seeks to build the world's most powerful, integrated scientific AI models for scientific discovery, applied energy deployment, and national security applications.

The Center for AI Policy provided responses to Question 3(a), which asked about open sourcing scientific and applied energy AI models, and Question 3(c), which asked about considerations for the DOE’s ongoing AI red-teaming and safety risks.

In response to these questions, we provide the following policy proposals: 

• ‍Safety testing: Consider both technical risks, which occur without malicious actors, and misuse risk when conducting safety testing.‍
• Partner organizations: Work with government agencies (e.g., the AI Safety Institute) and industry partners (e.g., METR) to tailor safety testing to a DOE context.‍
• Open sourcing: Adjust Know Your Customer (KYC) requirements and available model components based on the results of safety testing.

Read the full comment here.