Briefing
- Google DeepMind, Microsoft and xAI have signed agreements with CAISI allowing the US government to test frontier AI models before public release, confirmed by the National Institute of Standards and Technology.
- The Center for AI Standards and Innovation, part of the US Department of Commerce, announced the deals saying they are key to understanding capabilities of powerful AI models and protecting national security.
- The review process will vet models for cybersecurity, biosecurity and chemical weapons risks, according to The Guardian.
- The arrangements are framed around national security testing, distinct from broader product or safety regulation.
- The White House is separately studying an AI security executive order and considering an AI working group to formalise oversight and pre-release vetting processes, while also distancing itself from tighter AI regulation more broadly, according to reporting by the Federal News Network and Politico.
- The UK's AI Security Institute is also involved in evaluation efforts, according to a post on the official Microsoft Blog.
Analysis
- Voluntary pre-release access agreements with CAISI position Google DeepMind, Microsoft, and xAI favorably ahead of any formal AI executive order by demonstrating cooperative compliance. This reduces the probability of adversarial regulation targeting frontier model deployment timelines and creates an incumbency advantage: firms already inside the government testing framework set the procedural template that any formal rule must accommodate, raising barriers for later entrants like Meta and Amazon who are not party to these agreements.
Historical Context
- 2023-2024
The Biden administration's October 2023 AI Executive Order required frontier model developers to share safety test results with the government before public deployment. The Trump administration rescinded it, creating the regulatory vacuum that these voluntary CAISI agreements now fill, using the same pre-release access mechanism but under a national security rather than safety framing.
- 2002
Microsoft's antitrust consent decree required ongoing government access to interoperability information, establishing a precedent where voluntary cooperative compliance agreements with the government function as regulatory moats by setting compliance templates that smaller competitors cannot easily meet.
Connected Events
Further reading
See Indexa more often on Google
Mark Indexa as a preferred source — your Top Stories will surface more Indexa coverage.
