Anthropic breaks ranks on open-weight AI, refusing to sign the industry's anti-ban letter

When Nvidia, Microsoft, Meta, Google and, eventually, OpenAI signed an open letter urging Washington not to restrict open-weight AI models, Anthropic was conspicuously absent — fuelling a narrative that it wanted such models banned outright. On 27.07.2026, chief executive Dario Amodei rejected that framing directly, stating that open-weight models “that don't have dangerous capabilities are a public good” and that Anthropic has never called for banning them. His actual concerns lie elsewhere: he wants Washington to block sales of advanced chips and chipmaking equipment to China and crack down on smuggling, to curb industrial-scale distillation operations that let rival labs narrow the gap to the frontier without matching compute, and to mandate cyber, biological and alignment safety testing for the most capable models regardless of whether they are open, closed, or built in China. The dispute traces back to the shock release of Moonshot AI's Kimi K3, an open-weight Chinese model that approached US frontier performance at a fraction of the cost, which convinced much of Silicon Valley that restricting open models would only cede ground to Beijing. Anthropic has sharpened the China angle further by accusing Alibaba's Qwen lab of running the largest distillation campaign it has ever detected against Claude, using some 25,000 fake accounts across 29 million exchanges. The episode leaves Anthropic as the industry's lone holdout on a question that will shape the next round of US AI export and safety policy.