Anthropic CEO Amodei doubles down on open-weight risk stance while insisting he never called for a ban

{ "@context": "https://schema.org", "@type": "NewsArticle", "headline": "Anthropic CEO Amodei doubles down on open-weight risk stance while insisting he never called for a ban", "datePublished": "2026-07-28T14:06:42+02:00", "dateModified": "2026-07-28T18:15:24+02:00", "author": { "@type": "Person", "name": "Matthias Bastian" }, "publisher": { "@type": "Organization", "name": "The Decoder" }, "description": "Anthropic CEO Dario Amodei is once again warning about the risks of open AI models while insisting he has never called for a ban. He argues that authoritarian states like China could overtake the US and that open models could be misused for biological or cyberattacks. Critics say he's mostly trying to protect his own business from cheaper competition.", "articleBody": "Anthropic CEO Dario Amodei is sticking to his view that capable open-weight models carry significant risks, but he's not against open weights in principle.\r\n\r\nIn a\u00a0blog post, he stresses that \"Anthropic has never advocated for a ban on open-weights models,\" adding that models \"that don't have dangerous capabilities are a public good.\" A US ban on using Chinese models wouldn't fix anything either, \"because bad actors are unlikely to be legitimate US businesses,\" Amodei writes. \"It\u00a0would\u00a0protect US AI companies from competition, but that has never been my goal.\"\r\n\r\nAmodei says he has two \"nightmare scenarios\": authoritarian states, especially China, could build more powerful models than the US and use them to gain military superiority or surveillance capabilities. And powerful open models could be misused for cyber or bio attacks. AI labs know from their usage data\u00a0that such attempts are already happening with closed models as well.\r\n\r\nOpen weights are riskier, he argues, because they lack safeguards and a released model can't be recalled. While open models also empower defenders, Amodei sees a strong imbalance on biological threats, where sufficiently capable models could weaponize viruses quickly while building defenses would take years even in the best case.\r\n\r\nA study by the UK AI Security Institute found that\u00a0open-weight models trail frontier models by a few months on cybersecurity. Amodei has warned about the capabilities of his own models before, warnings that helped prompt a US ban, so he's at least consistent in his arguments.\r\nAmodei wants to slow China down and avoid worst-case scenarios\r\nAmodei is calling for stricter chip export controls\u00a0so China can't train more powerful models than the US without US chips. He wants action against\u00a0industrial distillation, which China reportedly uses to partly sidestep the chip restrictions and act as a \"free rider\" on US AI success. Critics counter that cutting off US chip sales to China mainly\u00a0fuels the growth of China's own chip industry\u00a0and that distillation is a legitimate and useful technique.\r\n\r\nAmodeli also wants mandatory safety tests for all powerful models, regardless of where they come from or how they're licensed. On that last point, Amodei sees a growing consensus, pointing to recent proposals that would exempt less powerful models from startups and research.\r\n\r\nAmodei's comments follow reports that US authorities are weighing steps to\u00a0make it harder for US companies to use Chinese open-weight models, along with a\u00a0letter signed by many major tech companies\u00a0opposing such moves.\r\n\r\nCritics accuse Anthropic of\u00a0talking down open-weight models\u00a0to protect its own business from cheaper competition. OpenAI is also said to still be lobbying against open weights,\u00a0even though the company signed the letter.", "isAccessibleForFree": "False", "hasPart": { "@type": "WebPageElement", "isAccessibleForFree": "False", "cssSelector": ".paywall-active" } }