US government to test new AI models for national security risks
US government to test new AI models for national security risks
Publish Date: 2026-05-05 14:41:00
Source Domain: www.usatoday.com
- Microsoft, Google, and Elon Musk’s xAI agreed to provide early access to new AI models for U.S. government national security testing.
- This initiative aims to assess the models’ potential capabilities and security risks before deployment.
- The agreement fulfills a 2025 pledge by the Trump administration to vet AI models for national security risks.
- The Center for AI Standards and Innovation at the Department of Commerce will evaluate the models and develop shared testing datasets and workflows.
- U.S. officials are increasingly concerned about national security risks posed by powerful AI systems, particularly in preventing cyberattacks and military misuse.
- This move builds on previous agreements with AI companies like OpenAI and Anthropic formed under the Biden administration.
- The Department of Commerce’s CAISI already has experience, having completed over 40 evaluations of cutting-edge AI models.
- The Pentagon has also formed agreements with AI companies to deploy advanced capabilities on its classified networks, although it did not disclose partnerships with companies like Anthropic currently embroiled in disputes over military AI use governance.
