US AI Safety Institute Renamed Center for AI Standards and Innovation
Summary
On June 4, 2025, Commerce Secretary Howard Lutnick announced the renaming of the US AI Safety Institute (AISI) to the Center for AI Standards and Innovation (CAISI). The institute remained housed within the National Institute of Standards and Technology (NIST) but underwent a mission reorientation — away from safety evaluations and toward facilitating commercial AI testing and standards development. Lutnick framed the prior AISI as a "censorship" vehicle that had impeded American AI competitiveness.
What Happened
The US AI Safety Institute had been established in November 2023 within NIST, modeled in part on the UK AISI announced at the same Bletchley Park summit. Its mandate covered voluntary safety evaluations of frontier AI models, development of evaluation methodologies, and coordination with international counterparts.
Commerce Secretary Lutnick's announcement marked a clean break from that framing. The new CAISI name dropped "safety" entirely, replacing it with "standards and innovation" — a phrase carrying explicitly commercial and technical rather than precautionary connotations. Lutnick's public statement characterized prior safety evaluations as having been used to block or delay model deployments, framing them as a mechanism for political interference in commercial AI development.
The reorganization also changed CAISI's operational priorities: the emphasis shifted to standards development to support US export competitiveness, facilitating private-sector AI testing at scale, and positioning the US as a standard-setter in international technical bodies such as ISO and the IEC. The safety evaluation partnerships AISI had built with frontier labs — including Anthropic, Google DeepMind, and OpenAI — were maintained but reframed as commercial testing facilitation rather than precautionary safety assessment.
Several senior AISI staff resigned in the weeks following the announcement. The institute's research agenda on evaluation methodology and red-teaming practices was significantly scaled back.
Why It Matters
The US AISI rename, coming four months after the UK's parallel rebrand, confirmed a transatlantic pattern: national AI safety bodies created in 2023 under the influence of frontier risk framing were being systematically repositioned by governments now prioritizing commercial competitiveness.
The loss of the word "safety" from both the UK and US institutes was not merely semantic. Naming shapes mission, budget, and staff recruitment. Research organizations whose formal mandate does not include safety are less likely to hire safety researchers, fund safety evaluations, or advocate for safety-oriented standards in international processes.
For the international AI safety institute network — which included counterparts in Japan, South Korea, Canada, Australia, and Singapore — the US and UK pivots created institutional uncertainty about the network's coherence and purpose. If the two founding members had moved away from the original mission, what was the network for?
The CAISI transition also highlighted a structural vulnerability of government AI safety programs: they depend on political will that can reverse between administrations. The AISI had been established by executive action and could be reorganized by the same means — no legislation was required to dismantle it.