AI Safety Deal: Google, Microsoft, xAI to Let US Review Models Before Release
Google, Microsoft and Elon Musk’s xAI have agreed to give the U.S. government early access to their latest artificial intelligence models so officials can review them for possible national security risks before public release.
The agreement marks another step in Washington’s growing effort to understand the power and risks of advanced AI systems. The reviews will be handled by the Center for AI Standards and Innovation, known as CAISI, which operates under the U.S. Department of Commerce.
Government to Review Advanced AI Before Release
Under the new arrangement, the companies will allow government experts to study their AI models before they are widely available. The goal is to test how these systems behave, what risks they may carry and whether they could be misused in areas such as cyberattacks, national defense or critical infrastructure.
CAISI said the agreements will support pre-deployment evaluations, targeted research and security testing. The agency has already completed more than 40 evaluations of advanced AI models, including some that have not yet been released to the public.
Why the Agreement Matters
The deal comes as U.S. officials and technology leaders grow more concerned about the fast development of frontier AI. These systems are becoming more powerful, and some experts worry they could help bad actors write malicious code, find software weaknesses or support other security threats.
Microsoft said it will work with U.S. government scientists to test AI systems for unexpected behavior and help develop shared testing methods. The company has also signed a similar agreement with the United Kingdom’s AI Security Institute.
OpenAI and Anthropic Already Involved
Google, Microsoft and xAI are not the first major AI firms to work with the U.S. government on model testing. OpenAI and Anthropic previously entered similar partnerships, and those agreements have now been updated to match the government’s latest AI policy direction.
The latest move shows that the government wants closer access to advanced AI tools before they reach the public. It also gives companies a way to show they are taking safety and national security concerns seriously.
AI Security Becomes a Bigger Priority
The rise of powerful AI models has created a difficult balance for governments. On one side, the U.S. wants to stay ahead in artificial intelligence innovation. On the other, officials want to prevent advanced models from creating new risks.
The agreement with Google, Microsoft and xAI suggests that AI safety reviews may become a more common part of the development process. Instead of waiting until a model is already public, government experts will be able to study some systems earlier and provide feedback.
For now, the program appears to be based on cooperation rather than strict regulation. But as AI tools become more capable, Washington’s role in reviewing them is likely to keep growing.
Apple’s Next Chapter: What to Expect From John Ternus’ Leadership
Apple’s Next Chapter: What to Expect From John Ternus’ Leadership The first thing to understand about John Ternus is that
Artemis II Mission: A Key Step Toward Designing a Moon Base
Artemis II Mission: A Key Step Toward Designing a Moon Base NASA’s Artemis II mission involves more than just transporting