Google DeepMind CEO Demis Hassabis' proposal for an industry-led AI safety regulator aims to curb frontier AI risks, but critics argue it lacks independent oversight, clear legal authority and safeguards against conflicts of interest.