Microsoft, Google, and xAI Open AI Models to U.S. Review

Microsoft, Google DeepMind, and Elon Musk’s xAI have agreed to give the U.S. government early access to their newest artificial intelligence models before they are released publicly. The agreement was announced by the Commerce Department’s Center for AI Standards and Innovation, known as CAISI. The goal is to let government researchers examine powerful AI systems ahead of public launch so they can better understand possible security risks, weaknesses, and national security concerns.

Under the new partnerships, CAISI will conduct pre-deployment evaluations and targeted research on AI models from Microsoft, Google DeepMind, and xAI. These reviews are meant to study the capabilities of frontier AI systems, especially the kinds of advanced models that could be used in cybersecurity, software development, planning, or other sensitive areas. CAISI said it has already completed more than 40 reviews of advanced AI models, and this new agreement expands its ability to test some of the most important systems being built by leading companies.

The move builds on earlier voluntary agreements made in 2024 with OpenAI and Anthropic, when CAISI was operating under the name U.S. Artificial Intelligence Safety Institute. Those earlier partnerships have since been renegotiated to better align with the Trump administration’s AI Action Plan. CAISI Director Chris Fall said independent and rigorous measurement science is important for understanding frontier AI and its national security implications. In simple terms, the government wants to know what these models can do before they become widely available.

The agreement comes at a time when concerns about AI safety and security are growing quickly. One major concern involves Anthropic’s newer AI model, Mythos, which reportedly raised alarms because of its ability to identify software flaws faster than most humans. While that kind of capability can be useful for cybersecurity, it can also create risks if powerful tools are misused. This is why the government is increasing its focus on testing models early, especially before they are released to the public or used in sensitive systems.

There are also signs that AI oversight could become more formal in the future. Reports suggest that the White House is considering an executive order that would bring together technology executives and government officials to oversee new AI models. For now, these partnerships remain voluntary, but they show that major AI companies are increasingly willing to work with government agencies on safety and security testing. The deals are also meant to improve information sharing, encourage voluntary product improvements, and help the government better understand AI capabilities and international competition.

The partnerships also reflect the growing connection between AI companies and national defense. The U.S. government is looking for ways to use advanced AI across areas like logistics, planning, cybersecurity, and classified systems without relying on only one provider. Google’s global affairs president, Kent Walker, said the company is prepared to support defense agencies, while Microsoft has also pointed to its responsible AI work. Together, these deals show that AI development is no longer just a private technology race. It is becoming a major national security issue, with governments and tech companies working more closely to decide how powerful AI should be tested, controlled, and deployed.

Works Cited

Iyke, Joy. “Microsoft (MSFT), Google (GOOGL) and xAI Grant U.S. Government First Access to AI Models.” TipRanks, 5 May 2026, www.tipranks.com/news/3273079-2. 

Roth, Emma. “Google, Microsoft, and xAI Will Allow the US Government to Review Their New AI Models.” The Verge, 5 May 2026, www.theverge.com/ai-artificial-intelligence/924017/google-microsoft-xai-government-review. 

Bordelon, Brendan. “US Government Expands Vetting of Frontier AI Models for Security Risks.” Politico, 5 May 2026, www.politico.com/news/2026/05/05/microsoft-xai-google-caisi-safety-testing-00906529. 

Microsoft Logo. Image. Reuters, 3 Jan. 2023, www.reuters.com/technology/ferrovial-build-data-centre-microsoft-madrid-cinco-dias-2023-01-03/.