Anthropic's Mythos Safety Report Shows It Can No Longer Fully Measure What It Built

OpenAI pauses development of Astra AI model over critical cybersecurity capability concerns