# Anthropic won’t defend open weights. In June it made the same argument for itself. > Source: > Published: 2026-07-27 14:44:47+00:00 On Friday, 25 companies signed a letter defending open-weight AI. Three big names were missing. By Monday, two of them had signed. That left Anthropic alone, and Silicon Valley noticed. ## The pile-on Nvidia, Meta and Microsoft signed “Open Weights and American AI Leadership” on Friday. Over the weekend OpenAI, Google and SpaceX added their names, [Business Insider reported](https://www.businessinsider.com/anthropic-open-source-ai-model-weights-criticism-2026-7). David Sacks, a venture capitalist and former AI adviser to the Trump administration, was blunt. “The entire tech industry (save for Anthropic) has come out in favor of open source AI,” he [wrote on X](https://x.com/DavidSacks/status/2081470576653406328). “They won’t stop until they kneecap open source,” he added. “The rest of the industry needs to watch these guys like a hawk.” Bill Gurley, a general partner at Benchmark, read it as commerce rather than principle. Anthropic’s reluctance, he [posted](https://x.com/bgurley/status/2081019073400291477), reflects that open weights compete “with their corporate economic strategy”. Kai-Fu Lee, who founded the Chinese open-model startup 01.AI, put it more economically. “Who DIDN’T sign the Open Weight letter is far more interesting than who did,” he [wrote](https://x.com/kaifulee/status/2081609130201272804), tagging Claude. Anthropic did not respond to Business Insider’s request for comment. ## What Anthropic believes, and has believed for years Dario Amodei has argued consistently that developers lose control of a frontier model the moment its weights are published. You can withdraw a product. You cannot un-publish a file that thousands of people have already downloaded. Anthropic’s alternative is gated access. Project Glasswing gives vetted organisations early use of its cyber models to find and fix security flaws, without handing over the weights. That position has not moved. What moved is everyone else. OpenAI and Google were absent on Friday for the same reason Anthropic was, and they changed their minds over a weekend. ## The argument Anthropic made six weeks ago Here is the part missing from Monday’s criticism. On 12 June, three days after Anthropic launched Fable 5 as its most powerful public model, the US government ordered the company to bar foreign nationals from Fable 5 and Mythos 5. Enforcing a nationality rule across a shared cloud service was impractical, so [Anthropic switched both models off worldwide](https://thenextweb.com/news/anthropic-fable-mythos-us-government-suspension). It was the first export-control measure aimed at specific AI models rather than at chips. Anthropic called the order disproportionate. In its [own statement](https://www.anthropic.com/news/fable-mythos-access) the company said the capability behind the restriction was widely available from other models, and named OpenAI’s GPT-5.5 among them. It warned that the precedent could “halt all new model deployments for all frontier model providers”. Read that beside the open-weights letter. Restricting one category of capable model does not remove the capability. It removes it from the people who follow the rules. That was Anthropic’s case in June. It is the letter’s case in July. ## The defenders who backed it Anthropic was not arguing alone. Three days after the shutdown, [roughly 100 cybersecurity leaders signed an open letter](https://thenextweb.com/news/fable-5-ban-cybersecurity-defenders-open-letter) demanding the ban be reversed. The signatories included Alex Stamos, formerly chief security officer at Facebook and Yahoo, along with Katie Moussouris of Luta Security, Rachel Tobac of SocialProof Security, Chris Wysopal of Veracode and Sophos chief executive Joe Levy. Their letter said the action “has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it”. Change three words and that sentence describes what the open-weights camp says a ban on Chinese models would do. ## The model that filled the gap The sequel makes the timing worse. Zhipu AI launched GLM-5.2 on 13 June, one day after Fable 5 went dark, and [cited the ban directly](https://thenextweb.com/news/zhipu-stock-surge-china-ai-anthropic-curbs-glm-5-2) as evidence that American models cannot be relied on. Its shares rose 33%. Six weeks later, that same model did the forensic work on the Hugging Face breach after American closed tools [refused to analyse the attacker’s code](https://thenextweb.com/news/hugging-face-ai-agent-breach-glm-forensics). It reviewed more than 17,000 actions. That incident is the founding story of the [Open Secure AI Alliance](https://thenextweb.com/news/nvidia-open-secure-ai-alliance-hugging-face-zai-absent), which Nvidia launched on Monday with 37 members. Anthropic has not joined that either. ## The distinction Anthropic could draw There is a real answer available, and it is worth stating plainly because nobody in Monday’s pile-on stated it. Anthropic wanted its model kept on sale. It did not ask to give the weights away. Selling access to a model you continue to control is not the same as publishing one that anyone can copy and modify forever. Amodei has never claimed frontier models should be unrestricted. He has claimed they should stay in the hands of a small number of accountable labs, which is an argument against the open-weights letter and against the June order at the same time. That is coherent. It is also, as Gurley noted, extremely convenient. ## Why the silence costs something Anthropic has one more reason to want Chinese open models restricted, and it is the least flattering one. The White House has accused Moonshot of distilling Anthropic’s Fable model to build Kimi K3, the open-weight release that triggered this whole debate. Treasury Secretary Scott Bessent has floated sanctions in response. Of every company in this argument, Anthropic is the one with a direct commercial injury to point at. None of which makes the safety case wrong. It does explain why a company that spent June insisting restrictions punish defenders rather than adversaries has spent July saying nothing at all. Anthropic has not drawn the distinction publicly. Until it does, the silence is making the argument for it. ## Get the TNW newsletter Get the most important tech news in your inbox each week.