US, Big Tech Firms Ink Pact for Pre-Release AI Model Performance, Security Checks

The US government has signed an agreement with three major tech companies to evaluate AI models before their release. The evaluation will be spearheaded by the Center for AI Standards and Innovation (CAISI), a division under the Department of Commerce. OpenAI and Anthropic had previously entered into similar agreements with the Commerce Department back in 2024.

Under the agreement, the three companies are required to provide AI models to CAISI in a state where safety measures that could potentially mitigate national security risks have been reduced or removed, allowing CAISI to measure these risks. The head of CAISI stated that independent and rigorous measurement is essential to understanding the national security implications of cutting-edge AI.

These measures come as the Trump administration reviewed new executive orders focused on cybersecurity and strengthened pre-release review procedures for AI models.