Anthropic claims it thwarted AI-based efforts at biological weapons—what does this mean for tech oversight?