Anthropic and OpenAI plan to integrate independent safety evaluators within their AI research teams. While this move is praised for offering rare access to internal processes, experts caution that effective oversight hinges on genuine independence, transparency, and the introduction of formal regulations.
Back