AI experts call for strict independent oversight on frontier AI models

More than 100 AI researchers are demanding genuine independence for third-party safety evaluators assessing frontier AI systems at OpenAI, Anthropic and other labs.