Staff at the federal center responsible for assessing advanced artificial intelligence systems have been reporting to the White House, The Associated Press reported Sunday. AP also reported that the center lacks a permanent leader and has stopped publishing its research. The report raises questions about who directs the center’s work and how much of it the public can see.
What the center is supposed to do
The National Institute of Standards and Technology’s description of the center says it works with industry on testing and research, helps develop voluntary standards, and leads unclassified evaluations of advanced AI capabilities that could pose national security risks. NIST identifies cybersecurity, biosecurity and chemical weapons among the risks the evaluations can examine.
That work is distinct from licensing an AI product for sale. NIST’s published remit describes voluntary agreements with developers and evaluators. Those agreements can give government researchers access to models for testing, while the center’s guidance can help other organizations compare how systems behave under similar conditions.
What to watch next
In February, NIST announced an initiative on standards for AI agents, systems that can carry out tasks using tools or other software. The agency said the effort would include research and guidance on interoperability and security. AP’s report concerns the center’s current reporting arrangements. The public NIST descriptions explain its remit but do not establish when the center will publish another evaluation.