Should governments, companies, or the open community regulate AI?

Mustafa Suleyman, in "The Coming Wave," argued that AI represents a technology so powerful that leaving it to market self-regulation is as reckless as letting pharmaceutical companies self-certify drug safety. He advocates for government-mandated licensing of frontier AI models above a capability threshold, mandatory safety testing before deployment, and international coordination modelled on nuclear non-proliferation frameworks. Geoffrey Hinton, since leaving Google in 2023, has called for government regulation with increasing urgency, arguing that AI companies are locked in a race where competitive pressure overrides safety caution. He has specifically advocated for mandatory interpretability research and compute governance — regulating the hardware that trains frontier models as a choke point. Yoshua Bengio has been the most active Turing Award winner in the policy space, co-authoring proposals for international AI governance and advising the Canadian government. He has called for mandatory pre-deployment safety evaluations for frontier models and public sector investment in safety research to counterbalance the private sector's profit motive. All three argue that industry self-regulation has a 100% historical failure rate when profits conflict with public safety.