US AI giant Anthropic alleges China rivals DeepSeek, Minimax and Moonshot AI are cheating
US AI giant Anthropic alleges China rivals DeepSeek, Minimax and Moonshot AI are cheating
https://www.cnn.com/2026/02/24/tech/anthropic-chinese-ai-distillation-intl-hnk
Publish Date: 2026-02-24 01:45:00
Source Domain: www.cnn.com
Certainly! Here are 4-8 key points from the article:
* Anthropic has accused three prominent Chinese AI labs—DeepSeek, Minimax, and Moonshot AI—of engaging in illegal practices to extract capabilities from its proprietary AI model, Claude.
* Anthropic alleged that these Chinese labs created over 24,000 fraudulent accounts and used over 16 million exchanges with the Claude model in an illegal process called distillation.
* Distillation is a common AI training method, but Anthropic and most proprietary AI model providers explicitly prohibit such practices, especially when the target model, like Claude, is not available in the country in question, China.
* The accusations echo similar allegations OpenAI made against DeepSeek and other Chinese AI firms for illegally distilling its ChatGPT model.
* The rapid advancements of Chinese AI models, such as DeepSeek, have raised questions about the effectiveness of U.S. tech and export controls.
* Anthropic expressed concern that illicitly distilled models may not have the same safety guardrails as original U.S.-developed models, potentially opening risks for national security and misuse in cybercrimes or bio-weapons.
* Anthropic warned that unauthorized distillation can enable authoritarian governments to misuse AI for offensive cyber operations, disinformation campaigns, and mass surveillance.
* The article underscores that the dependence on American models for advanced AI breakthroughs underscores the value of U.S. export controls in maintaining technological leadership in AI.
These points capture the core issues and concerns discussed in the article.
