Improving AI evaluations for a smarter future

Challenge: NIST’s mission is to develop and apply technology, measurement science, and standards to enhance productivity, facilitate trade, and improve the quality of life for U.S. citizens. As artificial intelligence (AI) technologies become increasingly prevalent, the agency is tasked with developing best practices for efficient risk-management and evaluating AI models to assess their dual-use capabilities and ensure the secure and effective adoption of AI across the federal government. The lack of these standards and reliable evaluation methods threatens U.S. government operations and public safety, and the absence of a centralized resource for AI testing and evaluation hinders adoption of AI in the U.S. government and the broader economy.

Approach: With TMF funding and support, NIST aims to accelerate efforts to create evaluation tools, test AI models and safeguards, develop AI best practices, and conduct relevant technical research. This work will help the federal government:

• Modernize faster by adopting AI responsibly
• Improve public-facing services
• Drive government efficiency and cost savings
• Enhance the understanding and management of rapidly emerging AI capabilities and risks