Anthropic's Amodei: no ban on open-weights models

What the post says

Anthropic CEO Dario Amodei published a post on 27 July 2026 setting out the company’s position on open-weights models, after reports that some US officials are considering banning US companies from using Chinese open-weights models. Many tech companies have signed a letter supporting open weights, and some commentators accused Anthropic of wanting a ban to protect its business. Amodei rejects that: “Anthropic has never advocated for a ban on open-weights models,” he writes, adding that models without dangerous capabilities are a public good, costing nothing beyond the compute needed to run them.

Two concerns, and why bans miss them

Amodei says protectionist bans would not address his two main national security worries, which he first laid out in his essay The Adolescence of Technology six months earlier. The primary one is that authoritarian governments — not only the Chinese Communist Party, though he calls it the most capable threat — build models more powerful than those of the US and use them for permanent military superiority or deep domestic repression. He notes that Vice President Vance warned in Paris last year about authoritarian regimes using AI to strengthen their military and intelligence capabilities. Whether such a model is released with open weights is irrelevant, he argues; the most dangerous one may be trained in secret and handed only to the People’s Liberation Army or the Ministry of State Security.

The secondary concern is misuse for cyberattacks or biological attacks, along with alignment problems. Open-weights models can carry higher risk than closed ones because guardrails are hard to apply and weights cannot be withdrawn once released — but banning their use by US businesses does nothing about it, since bad actors are unlikely to be legitimate US firms. “It would protect US AI companies from competition, but that has never been my goal,” he writes.

The three measures he backs

Instead, Amodei restates three policies he says he and Anthropic have consistently advocated. The first is to stop selling powerful chips and chipmaking equipment to China and to crack down on smuggling and workarounds: because China’s domestic production capacity is limited, it cannot build more powerful models than the US without American chips. The second is to crack down on industrial-scale distillation, which he says lets China build better models than its chip stock would allow and partly evade chip bans. The third is mandatory safety testing for all sufficiently capable models, open and closed, before release; he calls the idea close to consensus and welcomes recent moves by the US administration in that direction.

On the open letter

Amodei agrees with much of the open letter — open weights expand access to the AI economy and give customers more control — but disputes its assertions that open weights necessarily make safeguards easier to develop, or that broad access helps defenders more than attackers. Such questions, he writes, should be answered by rigorous pre-release testing rather than assumed.