Warning against over-restrictive AI policies that cede leadership to China

Big Tech's instinct to lock away its most advanced AI models is counterproductive and risks ceding the future of the technology to China. The cheap Chinese model from Moonshot is a wake-up call: secrecy does not prevent competitors from innovating; it only slows down the U.S. side. By keeping models closed, companies like OpenAI and Google are hampering the open research community that has driven AI breakthroughs historically. Open-weight models allow for independent audits, faster iteration, and wider access to benefits. The fear that Chinese labs will misuse open models is valid, but closing our own models does not stop them from building their own—they already have. In fact, by limiting access, we drive Chinese researchers to rely more on their own systems, accelerating their independence. Dario Amodei's concerns, while understandable, should not translate into a blanket policy of secrecy. The better response is to invest more aggressively in safety research and international agreements, not to hoard technology. History shows that open ecosystems win in the long run. If U.S. companies continue to lock away their best work, they may find themselves trailing behind Chinese competitors that embrace openness and collaboration.