Elon Musk said AI poses a very significant danger but questioned why laboratories such as Anthropic and OpenAI warn that their models could be catastrophic while seeking billions of dollars to scale them. Speaking at the All-In Summit, Musk cited the contradiction between warning of a 10% chance of human annihilation and asking investors about allocations in an IPO. He pointed to cyberattacks, possible future control of military systems and vulnerabilities in supposedly isolated systems that receive software updates through physical media. Musk proposed that AI companies give competitors API access to new models before release, allowing rival labs to identify and publicly disclose serious safety issues if developers fail to address them. He also supported open-sourcing the safety and testing apparatus. His proposal would supplement regulation with adversarial peer review, amid concerns that oversight can become a one-way ratchet and his preference for a measure that could win support from both the U.S. and China.