Anthropic has released a new safety framework for AI agents, a direct response to a wave of industry failures from Google, ...