Supporting Big Tech's cautious approach to releasing advanced AI models

Big Tech companies are right to be cautious about releasing their most powerful AI models openly. The arrival of a cheap Chinese model only reinforces the need for careful stewardship. While critics argue that secrecy stifles innovation, the reality is that releasing open-weight models to the public—especially those that can generate harmful content or be weaponized—poses unacceptable risks. The Chinese model, though impressive, also raises red flags: it could be used to spread disinformation or power autonomous weapons. Anthropic's Dario Amodei is not alone in fearing that Chinese labs, which operate under different regulatory and ethical standards, could accelerate the misuse of AI. By keeping their top models proprietary, U.S. firms can maintain control over how the technology is deployed, monitor its use, and fix flaws before they cause harm. Moreover, the competitive pressure from China actually justifies a slower, more deliberate approach. Racing to release models without adequate safeguards would repeat the mistakes of social media platforms, which prioritized speed over safety. Supporting this cautious policy is not about stifling progress; it is about ensuring that progress benefits society without catastrophic side effects.