Technology
Archived — This article has been archived. The information may be outdated.
Anthropic CEO calls China top AI security threat
NDTV··28 Jul
Anthropic co-founder and CEO Dario Amodei called for sanctions on China, describing it as the most capable authoritarian threat in artificial intelligence. In a blog post, Amodei said China may be distilling advanced American models to power open-weight systems such as Moonshot AI's Kimi K3. US officials had earlier alleged Moonshot distilled Anthropic's Fable model to build K3.
Prism
What It Means For You
- Developers using open-weight AI models should watch this debate, since it could shape which models remain freely available.
- Anyone concerned about AI safety should note Amodei's warning that biology-related misuse is harder to defend against than to cause.
- Businesses relying on Chinese open-weight models may face new uncertainty if the US moves toward sanctions or restrictions.
What's Happening
- Anthropic CEO Dario Amodei called China the most capable authoritarian AI threat and urged a crackdown on industrial-scale distillation.
- US officials had alleged China's Moonshot AI distilled Anthropic's Fable model to build its Kimi K3 system.
- China's commerce ministry said it would take necessary measures if sanctioned, alleging American firms have also distilled Chinese models.
The Technique at the Center of the Dispute
- Distillation lets a smaller model learn from a larger one's outputs, making it far cheaper than training from scratch.
- Amodei says distillation could bring China's AI frontier within months of the US despite chip export controls.
- About 200 Silicon Valley startups have urged Washington against banning open-weight access, warning it could hurt smaller developers.
all-news
