Anthropic CEO Dario Amodei said on Monday that he has "never advocated for a ban on open-weights models," rejecting accusations that his lab is trying to protect its closed-model business by lobbying against open-weight AI. In a new blog post he described models without dangerous capabilities as "a public good."
Why this matters
Anthropic has emerged as a prominent holdout from a recent industry push to defend open-weight AI. Major companies including Nvidia, Microsoft, Meta, Google and OpenAI, along with dozens of others, signed a letter urging Washington not to restrict the technology. The campaign gained momentum after the surprise debut of Kimi K3, a Chinese open-weight model that approached U.S. frontier performance at a small fraction of the cost and unsettled parts of Silicon Valley.
The Trump administration has also considered actions targeting open-source and open-weight models, including sanctions on Chinese labs accused of appropriating U.S. AI research through so-called "distillation."
Amodei’s position and reasoning
Amodei emphasized that he does not support a ban, while expressing serious national security concerns about the most capable systems. He argued that banning Chinese open models would do little to address the real threat, since "bad actors are unlikely to be legitimate US businesses." He acknowledged that a ban would protect U.S. labs like Anthropic from competition, but insisted "that has never been my goal."
Amodei is attempting to draw a narrow line: opposing an outright ban on open-weight models while advocating stricter controls on the most powerful AI systems. He concurred that open models can expand access, foster competition and increase customer control, but he rejected the claim advanced in the industry letter that open weights inherently make AI safer or give cyber defenders an edge over attackers.
Proposed alternative measures
Rather than a blanket prohibition — which he said "would not address my most serious national security concerns" — Amodei proposed three narrower policies:
- Tighter controls on the export of advanced chips and chipmaking equipment to authoritarian governments, including the Chinese Communist Party.
- A crackdown on "industrial-scale distillation," meaning the practice of training models on the outputs of more advanced systems.
- Mandatory safety testing for models that become sufficiently capable, regardless of whether they are open or closed.
Industry context and criticism
Anthropic drew criticism for declining to join the open-weight letter led by Nvidia CEO Jensen Huang. Google and OpenAI, two of Anthropic’s largest closed-model competitors, signed the letter over the weekend after initially abstaining; Anthropic has not. That absence fueled accusations — including from White House AI adviser David Sacks — that the company was using safety concerns as cover for protecting its own business model.
Bottom line
Dario Amodei does not want Washington to ban open-weight models outright. He wants policymakers to regulate the chokepoints that determine how powerful models can become — specifically chips, distillation practices and mandatory safety testing — rather than impose a blanket prohibition.
