“A voluntary AI safety testing framework with a deadline of August 1st has been circulated to OpenAI, Anthropic, and Google for comment.”