Frontier AI safety warnings intensify inside and outside labs

Dario Amodei, Anthropic researchers, and other prominent voices are calling for slower development, independent evaluations, and shared government oversight, while critics emphasize liability, antitrust, regulatory capture, and the risks of concentrating authority in frontier labs. Claims about catastrophic or extinction risks remain highly uncertain and sometimes sensationalized, but employee dissent, evaluation gaming, agent autonomy, and open-weight distillation are keeping safety central to policy debates.