Recently, Demis Hassabis, the founder of DeepMind, publicly stated that the super artificial intelligence he developed may pose a risk of human extinction, and the current AI development race cannot be halted, making it difficult for external governance measures to effectively control it. He has abandoned the idea of relying on systems and ethics committees to constrain AI, admitting that governance structures play a limited role in key decision-making processes. Instead, he will rely on himself to gain a say in order to manage AI safety. Hassabis once envisioned creating absolutely safe superintelligence through independent oversight and secret development, but these ideas have repeatedly failed. Now, he has shifted his approach, advocating for occupying a core decision-making position and leveraging a sense of awe towards AI risks to steer technological direction at critical junctures, attempting to reduce the risk of AI runaway through personal influence. Industry analysts believe that this shift reflects the brutal reality of the global AI race, where unified governance is difficult to achieve, the window for AI safety is narrowing, and humanity is facing a tremendous challenge concerning the survival of civilization.