The US government has reached agreements with Google DeepMind, Microsoft, and xAI to review early versions of their new artificial intelligence models before they are released to the public. The Center for AI Standards and Innovation (CAISI), a division of the US Department of Commerce, announced the deals on Tuesday, emphasizing that the review process is crucial for understanding the capabilities of advanced AI models and protecting national security.
CAISI stated that these collaborations will help the federal government scale its work in the public interest during a critical period. Chris Fall, CAISI director, highlighted that independent, rigorous measurement science is essential for understanding frontier AI and its national security implications. CAISI is designed to facilitate collaboration between the tech industry and the federal government in developing standards and assessing risks for commercial AI systems. The agreements focus primarily on identifying national security risks related to cybersecurity, biosecurity, and chemical weapons.
OpenAI and Anthropic previously signed similar deals with the Biden administration two years ago. CAISI noted that it has already completed over 40 such evaluations, including assessments of unreleased models. Developers often share AI models with reduced or removed safety guardrails to allow the government to thoroughly evaluate national security-related capabilities and risks.
The new agreements come amid growing concerns that the most powerful AI models could pose dangers if released publicly. AI safety experts, government officials, and tech companies fear that these models could enable hackers to exploit cybersecurity vulnerabilities on an unprecedented scale. For instance, Anthropic limited the rollout of its Mythos model to a few companies and launched Project Glasswing to unite tech companies in securing critical software.
Reports from the New York Times and Wall Street Journal on Monday indicated that the Trump administration is considering an executive order to create a government oversight process for AI tools. However, the administration has dismissed these reports as speculation. Google and xAI did not immediately respond to requests for comment. Meanwhile, Microsoft announced a similar agreement in the UK with the government-backed AI Security Institute, which also focuses on safe AI development.
In a blog post, Microsoft stated that while it regularly conducts various types of AI testing on its own, testing for national security and large-scale public safety risks must be a collaborative effort with governments.
