The federal government is significantly expanding its role in evaluating cutting-edge artificial intelligence systems before they reach the public, signing new agreements with Google DeepMind, Microsoft, and xAI that grant officials direct access to test powerful models, according to a Commerce Department announcement.
The expanded partnerships mark a critical juncture in how the nation's regulatory apparatus addresses the rapid development of frontier AI technologies—systems with capabilities that remain poorly understood even by their creators. Under the new agreements, the government's Center for AI Standards and Innovation (CAISI) will conduct pre-deployment evaluations and targeted research to assess frontier AI capabilities and advance the state of AI security before models are released to the public.
Government's Expanding Role
CASI director Chris Fall emphasized the stakes of this expanded oversight. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," Fall said. "These expanded industry collaborations help us scale our work in the public interest at a critical moment."
The announcement comes as the Trump administration considers further strengthening its hand through potential executive action on cybersecurity and pre-clearance requirements for new models. The Commerce Department indicated that the new agreements with Google, Microsoft, and xAI build on partnerships with Anthropic and OpenAI that were first launched 2 years ago, with those earlier deals now renegotiated to reflect CAISI's directives and the administration's broader AI action plan.
Institutional Evolution and Continuity
The government's AI testing apparatus has undergone significant transformation. A 2023 executive order established the AI Safety Institute under the Biden administration, which was subsequently re-named under the Trump administration as CAISI. Reports indicated that CAISI underwent significant changes at the beginning of Trump's term and was expected to pivot from AI safety to AI acceleration. Yet the institute has continued conducting rigorous AI testing and evaluations, including publishing an evaluation of China's DeepSeek and soliciting public comment on secure deployment of AI agents.
The institutional continuity reflects a bipartisan recognition that frontier AI development requires some form of government engagement—though the specific emphasis and direction have shifted. Leadership changes have marked the transition: Chris Fall was recently announced as director of CAISI after former Anthropic staffer Collin Burns was reportedly pushed out after just four days in the role.
Why This Matters:
This expansion of government oversight addresses a fundamental democratic accountability gap: the development of increasingly powerful AI systems by private corporations with limited transparency or public input. The agreements represent an effort to ensure that frontier AI capabilities are evaluated for security risks and societal impacts before deployment, rather than after potential harms have already occurred. By requiring pre-deployment testing and post-release assessments, the government is attempting to assert democratic control over technologies that could have far-reaching consequences for workers, consumers, and national security. The continuity of testing and evaluation work across administrations suggests recognition that some form of institutional oversight is necessary to manage the risks posed by rapidly advancing AI systems—a principle that reflects the center-left view that market forces alone cannot adequately protect the public interest in critical emerging technologies.