News-Google-MS-xAI-CAISI-Early-Model-Access-May-2026

Google DeepMind, Microsoft, and xAI have agreed to give the U.S. government early access to their frontier AI models for evaluation, the U.S. Commerce Department's Center for AI Standards and Innovation (CAISI) announced on May 5, 2026. The three companies join OpenAI and Anthropic, which had previously granted CAISI pre-deployment access for security assessments and capability evaluations.

The expansion marks a significant step in the government-industry collaboration on AI safety, bringing the total number of participating frontier AI labs to five. CAISI, established under the Commerce Department, conducts pre-deployment evaluations to assess the capabilities of advanced AI systems before public release. The center aims to identify security vulnerabilities and potential misuse risks in next-generation models.

The announcement follows a May 4, 2026 New York Times op-ed by former Trump and Biden AI advisers Dean Ball and Ben Buchanan, who called for bipartisan action on AI security risks including mandatory safety audits for frontier AI models. It also comes as the Trump administration is reportedly considering an executive order to form an AI working group focused on model vetting procedures.

Significance

With five major AI firms now voluntarily participating in pre-deployment evaluations, CAISI's program represents a voluntary cooperative model for AI governance that contrasts with mandatory regulatory frameworks adopted in other jurisdictions such as the European Union. The inclusion of xAI, founded by Elon Musk, is notable given Musk's ongoing litigation against OpenAI (Musk v. Altman) over AI safety and corporate governance concerns.

See Also

• Ball & Buchanan Bipartisan AI Security Appeal
• May 4, 2026 Daily Digest (Trump AI Working Group EO discussion)
• Musk v. Altman Case Page