Anthropic Says It Does Not Support a Ban on Open-Weight AI Models

Anthropic CEO Dario Amodei said Anthropic does not support a categorical ban on open-weight AI models, while warning that model releases can affect U.S.-China competition in advanced AI.

Amodei clarifies Anthropic’s position

In a post published by Anthropic titled “Our position on open-weights models,” Amodei wrote that Anthropic has “never advocated for a ban” on open-weight models as a category. The statement responds to criticism that the company has sought to restrict open releases broadly.

Amodei’s distinction is important: he defended the legitimacy of open-weight AI development while arguing that not all models carry the same risks. In Anthropic’s framing, the policy question is not whether open-weight models should exist, but how governments should handle increasingly capable systems that may be useful to strategic competitors or unsafe actors.

Axios reported that Amodei rejected claims that Anthropic wants to ban open-weight AI and instead called for a combination of tighter chip controls, measures against industrial-scale distillation, and safety testing. Bloomberg Law similarly reported that Anthropic said it never sought an open-weights ban and supports restrictions on sales of powerful chips and chipmaking equipment to China.

The China concern

Amodei’s post argues that the Chinese Communist Party is the most capable authoritarian actor that could plausibly build AI systems more powerful than those in the United States. He wrote that U.S. policy should consider how open model releases interact with export controls aimed at limiting China’s access to advanced compute.

The concern is that open-weight releases can transmit technical progress beyond the organizations that trained the original models. Unlike closed systems accessed only through an API, open-weight models can be downloaded, modified, fine-tuned, and studied by outside developers. That openness can support academic research, startups, and independent builders, but it can also make capabilities available across borders.

TechCrunch reported that Amodei’s comments were aimed at separating Anthropic’s support for open-weight development in general from its worries about frontier-level releases and geopolitical risk. The company’s position places more emphasis on export controls and safety evaluation than on a blanket prohibition.

Open weights remain a contested policy issue

The debate comes as major AI developers and governments weigh the benefits of open models against the potential risks of rapid capability diffusion. Supporters argue that open weights improve transparency, competition, independent auditing, and access for researchers and smaller companies. Critics argue that highly capable open models may be harder to control once released, because the original developer cannot reliably limit downstream use.

Anthropic’s latest statement does not resolve that tension, but it narrows the company’s stated position. Amodei is not arguing that all open-weight models should be stopped. He is arguing that policymakers should assess the most capable systems in the context of national security, safety testing, and the effectiveness of chip export controls.

That approach is likely to remain controversial. Open-source and open-weight advocates often warn that restrictions could entrench the largest AI companies by making it harder for outside researchers and smaller firms to compete. Security-focused policymakers, meanwhile, are increasingly concerned that frontier model capabilities may spread faster than enforcement tools can adapt.

What to watch

The practical question is how governments define which models require additional scrutiny. A policy that treats every open release the same would be difficult to justify and could chill lower-risk research. A policy focused only on the most capable models would need clear thresholds, credible evaluation methods, and safeguards against becoming a de facto barrier to competition.

Amodei’s comments suggest Anthropic wants the debate to center on those thresholds rather than on a simple open-versus-closed divide. Based on the company’s statement and reporting from Axios, Bloomberg Law, and TechCrunch, Anthropic’s public position is that open-weight AI should not be banned as a category, but that frontier releases should be considered alongside chip controls, distillation risks, and safety testing.