🚨 Google DeepMind wants frontier AI models reviewed before they are released

They have proposed a new US-led global AI standards body:

• Frontier labs would submit models up to 30 days before release • Both open and closed models would be evaluated • Testing would examine cyber, biological and national security risks • The body would be industry funded but accountable to the US government • It could coordinate an industry wide slowdown if a model was considered too dangerous

Should the US Government have the power to delay a major AI model launch or refuse its launch to the public?