In a blog post released Monday, Anthropic CEO Dario Amodei addressed what he described as misunderstandings about the company’s position on open-weight artificial intelligence models, saying Anthropic has “never advocated for a ban on open-weight models.”

Amodei rejected the idea that the company supports government prohibition of such systems. He said that AI models without dangerous capabilities are “a public good,” and acknowledged that a ban could shield Anthropic from competition, but said that was not the company’s motivation.

ANTHROPIC CEO: OPEN SOURCE AI IS GETTING DANGEROUS

Anthropic CEO Dario Amodei told lawmakers that open-source AI is moving down a “very dangerous path.”

His warns that once powerful models are released openly, companies lose the ability to monitor misuse, revoke access, or… pic.twitter.com/099ptGEEP4

— Coin Bureau (@coinbureau) June 28, 2026

Anthropic’s policy proposals

Rather than calling for a ban, Amodei laid out three policy recommendations.

First, he called for tighter export controls on advanced semiconductor chips and manufacturing equipment sent to authoritarian countries, with particular focus on China. He said China’s limited domestic semiconductor manufacturing capacity means it cannot build AI systems that surpass U.S. capabilities without access to American chip technology, making export controls the most practical policy tool.

Second, he urged restrictions on large-scale AI distillation, the process of training smaller models using outputs generated by more capable systems.

Third, he called for mandatory safety evaluations for all sufficiently advanced AI models, whether they are open or proprietary.

Amodei said Anthropic recognizes that open-weight models can broaden access to AI and increase competition. However, he said the company does not accept the view that open-weight systems are inherently safer or that they necessarily provide defensive advantages against malicious actors. That framing places Anthropic in a broader policy fight that now extends beyond model release decisions to the infrastructure and methods used to develop advanced AI, especially as governments weigh how to limit military or surveillance applications without constraining wider commercial use.

Technology industry support for open-weight models

The policy debate has split major technology companies. Last week, Nvidia CEO Jensen Huang made his first post on X to support open-source AI and signed a letter arguing that open models accelerate innovation and strengthen national AI leadership.

Google CEO Sundar Pichai also signed the letter. Other signatories included Microsoft, Advanced Micro Devices, Cisco Systems, Cloudflare, GitHub, and many others. Anthropic and Amazon did not sign.

OpenAI, which typically operates closed systems, also signed the letter supporting open-weight models.

The discussion intensified after Chinese company Moonshot AI released Kimi K3, an open-weight model that reportedly delivers near-frontier performance at lower cost than U.S. rivals.

Reports have also said the Trump administration has considered measures aimed at open-source AI development, including possible sanctions against Chinese research institutions accused of using distillation methods to replicate American AI systems.

Amodei said banning open-weight models would not address his main national security concerns, arguing that semiconductor export restrictions are a more targeted tool than broad model bans.