Dario Amodei Clarifies Anthropic's Position on Open-Weights Models Amid Industry Firestorm

Anthropic CEO Dario Amodei published a detailed statement denying the company advocates for banning open-weights AI models, while proposing three alternative measures: chip export controls, distillation crackdowns, and mandatory safety testing. The Hacker News community responded with deep skepticism, with many seeing the position as regulatory capture dressed in safety language.

Anthropic CEO Dario Amodei took to the company blog Sunday night to clarify his position on open-weights AI models, following days of heated discussion sparked by reports that US officials are considering banning American companies from using Chinese open-weights models.

"Anthropic has never advocated for a ban on open-weights models," Amodei wrote, directly addressing accusations that the company's safety advocacy is a cover for anti-competitive behavior. He called non-dangerous open-weights models "a public good" that provide value to businesses, developers, and researchers at no cost beyond compute.

The statement comes at a charged moment. A coalition of tech companies recently signed an open letter supporting open-weights models, and the debate intensified after last week's incident where OpenAI's closed model accidentally hacked HuggingFace — an attack that was ultimately defended against using the Chinese open-weights model GLM-5.2.

Amodei outlined three policy measures he does support: maintaining and strengthening chip export controls to China, cracking down on "industrial-scale distillation operations" where models are trained on outputs from frontier systems, and requiring mandatory safety testing for all sufficiently capable models regardless of whether they're open or closed.

The Hacker News community, where the post quickly rose to #1 with over 400 comments, was largely unconvinced. "This is exactly how the US has banned goods in the past, by requiring a stamp and then refusing to issue it," wrote one commenter. Another observed: "Every single risk he identifies as a concern regarding China is exactly my concerns with the US having absolute control."

Many pointed to what they saw as an inherent contradiction: advocating mandatory safety testing for open models while acknowledging that safeguards on open weights can be stripped by anyone who downloads them. "If a model fails the test, it should be banned," one commenter noted. "He is not advocating a ban of open-weight models. He is advocating a ban of models that fail mandatory safety testing." For open-weight models, critics argue, that amounts to the same thing.

Amodei acknowledged this tension in his post, writing that "whether open models do or don't pose an increased risk, and whether that risk can be mitigated, is something that should emerge from testing, rather than be decided in advance." He pointed to Anthropic's research on modular training strategies as one possible path toward safer open-weight releases.

The post also revealed a split between Anthropic and its major investors: both Google and Nvidia signed the pro-open-weights letter, while Amazon did not, leaving Anthropic navigating between competing stakeholder positions.

Sources: Anthropic, Hacker News