US to Test Google, Microsoft, xAI AI Models for Safety

The US Department of Commerce will now safety test new artificial intelligence models from Google, Microsoft, and xAI before they are publicly released. These tech companies have agreed to voluntarily submit their AI models for evaluation through Commerce's Center for AI Standards and Innovation, known as CAISI. This expands on previous agreements with other AI companies. CAISI's director, Chris Fall, states these collaborations are crucial for scaling public interest work. The evaluations will cover testing, research, and best practice development for commercial AI systems. For example, Google's Gemini, Microsoft's CoPilot, and xAI's Grok will undergo this scrutiny. CAISI has already conducted 40 previous evaluations, including some for unreleased, state-of-the-art models. This shift indicates a growing focus on AI safety and national security. This matters because it aims to ensure new AI tools are safe and secure before they reach the public.