US AI Safety Institute becomes the Center for AI Standards and Innovation

The Center for AI Standards and Innovation (CAISI) is the new name for the agency inside the U.S. Department of Commerce that had been known as the U.S. AI Safety Institute. Commerce Secretary Howard Lutnick announced the change on 3 June 2025.^[1]

Background

The institute sits within the National Institute of Standards and Technology (NIST), the Commerce Department's standards body, and the announcement kept it there: CAISI is to continue operating inside NIST and to coordinate with other Commerce bureaus, including the Bureau of Industry and Security.^[1]

What was announced

Lutnick said the reorganisation is meant to ensure the department uses its scientific and industrial expertise to evaluate the capabilities of rapidly developing AI systems and to identify vulnerabilities in systems built in the United States and abroad.^[1] In the same statement he said that "censorship and regulations have been used under the guise of national security", and that innovators would no longer be limited by such standards.^[1]

What the centre is meant to do

According to the announcement, CAISI is to serve as industry's primary point of contact within the U.S. government for testing and for collaborative research on commercial AI systems.^[1]

Working with NIST organisations, it is to develop guidelines and best practices for measuring and improving the security of AI systems, and to help industry develop voluntary standards.^[1]

It is also to establish voluntary agreements with private-sector AI developers and evaluators, and to lead unclassified evaluations of AI capabilities that may pose risks to national security.^[1] In those evaluations the centre is to concentrate on demonstrable risks, with cybersecurity, biosecurity and chemical weapons named in the announcement.^[1]

Beyond testing individual systems, CAISI is to assess the capabilities of U.S. and adversary AI systems, the adoption of foreign AI systems, and the state of international AI competition.^[1]

Who it works with

The announcement lists coordination with the Department of Defense, the Department of Energy, the Department of Homeland Security, the Office of Science and Technology Policy and the intelligence community.^[1] Within Commerce it is to coordinate with other bureaus, including the Bureau of Industry and Security.^[1]

What the change amounts to

Read together, the elements of the announcement describe a shift in emphasis rather than a move out of government standards work. The agency keeps its home in NIST and keeps a testing and evaluation mission, but the instruments it is told to use are voluntary: voluntary agreements with developers and evaluators, and voluntary standards written with industry rather than mandatory rules.^[1] Its stated role toward companies is that of a single entry point for testing and joint research on commercial systems.^[1]

The scope of the evaluations is also set out explicitly. They are to be unclassified, they are to cover systems built abroad as well as in the United States, and they are to be aimed at risks the announcement calls demonstrable — cybersecurity, biosecurity and chemical weapons.^[1] Alongside the technical work, the centre is given a watching brief on how foreign AI systems are being adopted and on the state of international AI competition.^[1]

The name carries the framing: standards and innovation on one side, safety on the other. Lutnick's statement ties the reorganisation to a rejection of what he described as regulation and censorship imposed in the name of national security, and to a promise that innovators would not be limited by such standards.^[1] What the announcement does not do is set out budgets, staffing or timetables; it describes the mandate and the partners, and leaves the operating details to the centre inside NIST.^[1]