Multiple prominent tech and security figures compare the regulation of advanced artificial intelligence to controlling nuclear weapons, arguing that powerful AI systems could pose difficult-to-contain risks. Elon Musk says AI is “far more dangerous than nukes.” Anthropic CEO Dario Amodei describes the strongest AI models as “weaponisable nuclear materials.” OpenAI CEO Sam Altman points to the International Atomic Energy Agency (IAEA) as a possible reference point for oversight.
Foreign Affairs and Times of India examine why the analogy is appealing yet potentially limited. Both note that some leaders reach for the nuclear comparison when discussing policing and governance mechanisms. The argument centers on whether existing nuclear-era institutions and safeguards can translate to AI, which may be developed and deployed differently from nuclear materials. While supporters of the analogy look to the IAEA model, critics question whether AI’s technical and operational characteristics make control as durable as nuclear arms regulation.
Times of India frames the US–China AI competition as a test case for how hard such control could be, whereas Foreign Affairs focuses on the broader “trouble” with using nuclear lessons for AI governance. Both discuss the same key public references but differ in emphasis on the analogy’s usefulness.
