NIST seeks contractors to build AI testing datasets for security and safety verification
Gettyimages.com/ tanit boonruen
By
Nick Wakeman
The National Institute of Standards and Technology wants curated datasets to test systems across cybersecurity, criminal misuse prevention and language translation capabilities.
The National Institute of Standards and Technology is in the early stages of developing a contract to access multiple data sets to test and verify artificial intelligence systems.
NIST’s U.S. Center for AI Standards and Innovation wants to test, evaluate and verify AI systems for reliability and robustness. The agency wants a contractor or multiple contractors that can design and curate datasets to test the AI systems in order to accomplish that goal, according to a sources sought notice posted Tuesday.
A sources sought notice is often issued for market research purposes before an agency drafts a solicitation. In this case, NIST is trying to determine whether there are commercial sources of datasets that can meet its needs.
Some of the common use cases NIST wants to test include security such as cyber, chemical and biological; criminal misuse of AI; software engineering; AI-human interaction; AI research and development; and agentic tasks.
NIST is looking to collect information on the size of the dataset, domain areas, tasks to be represented and formats.
NIST also wants translation and localization services covering 19 languages that can serve as either source or target languages, meaning contractors must be able to translate from any of these languages to any other on the list.
Responses are due Oct. 10.