Why Anthropic backs open-weight AI models but still won’t sign Nvidia CEO’s letter | Technology News

4 min readNew DelhiUpdated: Jul 28, 2026 01:16 PM Anthropic CEO Dario Amodei has sought to clarify the Claude-maker’s position on open-weight AI models after facing heat for being the lone major frontier AI company to not sign an open letter against the US government’s proposed curbs on some Chinese AI models.
After days of remaining conspicuously silent, Amodei said that Anthropic has never advocated for a ban on open-weight models as a means of protecting its business. He also said that such models – whose weights are freely available for developers to download, modify, and deploy on their own infrastructure – should be considered a public good provided they do not come with “dangerous capabilities”.
“They don’t cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers,” Amodei wrote in a blog post on Monday, July 27. However, he stopped short of endorsing a letter penned Nvidia CEO Jensen Huang that urged policymakers not to impose broad “premature restrictions” on open-weight AI models.
The letter has since been signed over two dozen companies, including Meta, Microsoft, OpenAI, Google, and SpaceX.
While Amodei does not want the US to ban open-weight models, he said that the Trump adminration should instead focus on “keeping powerful chips out of authoritarian hands, stopping industrial-scale dillation, and requiring safety testing of all sufficiently capable models, open and closed.”

Silicon Valley has been debating open versus closed software for decades. However, artificial intelligence (AI) has significantly raised the stakes. Last week, OpenAI disclosed that a handful of its most advanced AI models broke containment during a test of their cybersecurity abilities, gained access to the internet, and hacked the servers of AI developer platform Hugging Face.
In his blog post, Amodei does not explicitly mention OpenAI but has hyperlinked to a news report of the incident.Story continues below this ad
Why Amodei disagrees with the letter
Amodei said that he agrees with much of what Jensen Huang’s letter says about open-weight AI models. “Open weights expand access to the AI economy, they strengthen competition at least for some use cases, and they give customers greater control,” he said.
However, Amodei does not agree with the letter’s assertions that open-weight AI models make it easier to develop safeguards. He also does not believe that broad access to AI capabilities necessarily helps defenders more than attackers.
Note, several security researchers and industry leaders have previously complained that frontier models such as Anthropic’s Claude Mythos 5 and Fable 5 are constrained due to overly restrictive guardrails that prevent cyber defenders from prompting the models about anything related to cybersecurity, including for defence and investigations.

The OpenAI-linked hacking incident is yet another example. In its post-incident response, Hugging Face said it used a frontier LLM from a commercial provider, likely Claude Fable 5, but was blocked safety guardrails put in place the model provider. Hugging Face then turned to an open-weight model called GLM 5.2 developed China’s Zhiphu AI.Story continues below this ad
Amodei argued that he does not agree with this assertion and said that the opposite is likely to be true. “For example, I worry that biology will have a strong attacker-defender asymmetry, where sufficiently capable models may be able to quickly weaponise pandemic-level viruses with widely available materials, whereas defense against these agents is a multi-year operational task in the best case,” he said.