Google, Microsoft, and xAI have entered into an agreement to provide the US government with pre-release access to their frontier AI models for vetting purposes. This development was announced by the Commerce Department’s Center for AI Standards and Innovation, highlighting its role in overseeing these collaborations.
“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” stated Chris Fall, the center’s director, emphasizing the importance of these collaborations in scaling their work for public interest.
This move comes in the context of ongoing discussions around executive orders and AI oversight.
US Government Pre-Release AI Model Vetting
Google, Microsoft, and xAI have agreed to provide the US government with pre-release access to their frontier AI models, a significant development announced by the Commerce Department’s Center for AI Standards and Innovation. These agreements were made ahead of the potential public release of the models. This initiative follows reports of the Trump administration considering an executive order to establish a working group tasked with evaluating advanced AI models before their public deployment. Such a move signifies a shift from earlier policy positions held by the Trump administration regarding AI regulation.
The New York Times reported that White House officials have engaged in discussions with executives from Anthropic, Google, and OpenAI regarding oversight plans for AI technologies. These discussions are part of broader efforts to ensure that advanced AI models are adequately reviewed, addressing both security and ethical concerns before they reach the general public. The proposed executive order would represent a reversal of the previous administration’s less regulatory approach to AI development and deployment.
Anthropic’s AI model, Claude Mythos, is recognized for its strong cybersecurity capabilities, particularly in identifying vulnerabilities. According to Mozilla, Mythos assisted in patching 271 vulnerabilities in Firefox, underscoring its effectiveness in enhancing software security. Despite its capabilities, there is skepticism within the user community about a wide release of Claude Mythos by June 30, with only a 13% chance estimated.
In February 2026, Anthropic faced a contract dispute with the Trump administration after it refused to grant unrestricted access to its AI models. This refusal led to Defense Secretary Pete Hegseth designating Anthropic as a supply chain risk to national security. As litigation on this matter continues, a federal appeals court declined to pause this designation. Amid these developments, Axios reported that the White House is considering whether to resume its partnership with Anthropic. These events highlight the complex interplay between technological innovation, national security concerns, and government-industry relations.
The Commerce Department’s Center for AI Standards and Innovation has secured agreements with major AI firms to provide pre-release access to frontier models for government vetting, reflecting a public–private collaboration on model assessment. This effort sits alongside discussions of potential executive actions to formalize oversight and includes separate controversies around Anthropic and its Claude Mythos model related to cybersecurity findings and access disputes, which have affected government partnerships.
