Anthropic CEO Dario Amodei has responded to criticism over the company's absence from a letter signed by dozens of technology firms urging the White House not to ban Chinese open-weight AI models.

In a blog post, Amodei said Anthropic does not support a blanket ban on open-weight models, emphasizing that the company has "never advocated for a ban."

He acknowledged that open-weight models provide significant benefits for developers, researchers and businesses because they are generally less expensive to run and can often be deployed locally.

What Are Open-Weight AI Models?

Open-weight AI models make their trained model parameters (or "weights") available, allowing developers to run and fine-tune them on their own hardware.

Compared with proprietary or closed-weight models, they typically offer:

• Lower deployment costs.
• Greater flexibility for developers.
• Local deployment without relying entirely on cloud services.
• Faster research and experimentation.

Examples frequently discussed in the current debate include:

• Kimi K3
• DeepSeek models.
• GLM 5.2.

Why Is Anthropic Concerned?

Although Amodei supports open-weight models in principle, he outlined concerns about how increasingly capable AI systems could be misused.

Concerns About China

Amodei argued that China could use advanced AI capabilities for purposes such as:

• Military applications.
• Cyber operations.
• Expanding state surveillance.
• Enhancing domestic censorship.

These remarks reflect Anthropic's assessment of potential national security risks. They represent the company's viewpoint and are part of a broader policy debate; different governments, companies and experts hold differing views on the risks and appropriate responses.

Risks Beyond Geopolitics

Amodei also argued that powerful AI models—particularly those with openly available weights—could be more difficult to control once released.

According to him, advanced AI could potentially be misused for:

• Sophisticated cyberattacks.
• Development of harmful biological capabilities.
• Other malicious activities if adequate safeguards are absent.

He said that once model weights are publicly released, they cannot easily be withdrawn or modified with new guardrails.

Anthropic's Policy Recommendations

Rather than banning Chinese open-weight AI models, Amodei proposed several policy measures.

1. Stronger Export Controls

He called for tighter restrictions on exports of advanced AI chips and semiconductor manufacturing equipment to China, along with stronger enforcement against alleged smuggling and other methods used to obtain restricted hardware.

2. Address AI Model Distillation

Amodei urged policymakers to tackle industrial-scale AI model distillation—the practice of training smaller models using outputs from more capable systems.

Anthropic has previously alleged that some Chinese AI developers have used distillation techniques to narrow the performance gap with leading frontier models.

3. Mandatory AI Safety Testing

Amodei argued that all frontier AI models, whether open-weight or closed-weight, should undergo rigorous safety evaluations before public release.

He suggested testing should include assessments of:

• Cybersecurity risks.
• Biological misuse risks.
• Model alignment and safety.

Industry Divided Over AI Policy

The discussion has highlighted differing views across the AI industry.

According to reports, more than 50 technology companies, including OpenAI, supported a letter urging the US government not to prohibit Chinese open-weight models.

Industry leaders such as Jensen Huang and Satya Nadella have argued that open AI systems encourage broader innovation, accessibility and competition.

Anthropic did not sign the letter, leading some critics—including former White House AI adviser David Sacks—to question the company's position.

Anthropic maintains that its concerns are focused on AI safety and national security rather than opposition to open-weight models.

Growing Competition from Chinese AI Labs

The policy debate comes as Chinese AI companies continue releasing increasingly capable models at lower costs.

Recent models from companies including Moonshot AI, DeepSeek and Z.ai have attracted attention for their performance on industry benchmarks and competitive pricing, intensifying global competition in the AI sector.

Why This Matters

The debate over open-weight AI models reflects broader questions about balancing innovation, security and global competition. As AI capabilities advance rapidly, policymakers and technology companies are weighing how to encourage research and accessibility while reducing risks related to misuse, cybersecurity and geopolitical competition. The differing positions among major AI firms underscore that there is no industry-wide consensus on the best regulatory approach.

Frequently Asked Questions (FAQs)

Did Anthropic call for banning open-weight AI models?

No. CEO Dario Amodei said Anthropic has not advocated a blanket ban on open-weight AI models.

Why is Anthropic concerned about China?

Amodei said he is concerned that increasingly powerful AI could be used for military, surveillance or other harmful purposes if developed without sufficient safeguards. These are Anthropic's stated concerns and are part of an ongoing policy debate.

What are open-weight AI models?

Open-weight models make their trained parameters available, allowing developers to run and adapt them on their own hardware.

What solutions did Dario Amodei propose?

He called for stronger AI chip export controls, action against industrial-scale AI model distillation and mandatory safety testing for advanced AI models before release.

Why is this debate important?

The discussion could influence future AI regulation, international technology competition and how advanced AI models are developed, distributed and governed.