Anthropic accuses DeepSeek and 2 other Chinese AI models of stealing its data, people on the Internet say it’s justified
Anthropic has accused three Chinese AI firms, including DeepSeek, of distilling data from its cloud chatbots to train and improve their own AI models. Surprisingly, instead of getting support, Anthropic got disappointment from people as they had accused the AI firm of doing the same thing earlier also.
This tension and cold war between Chinese AI companies and their American counterparts has been going on for some time. American giant Anthropic is at the forefront of cloud AI. On Tuesday, the company opened a new front in this war as it accused three major Chinese AI companies, DeepSeek, MiniMax and Moonshot AI, of stealing data from the cloud to train their AI models.
❮❯
Anthropic said in a blog post that Chinese companies used thousands of accounts to interact with the cloud amounting to millions of tokens. Information gained from these conversations was used to improve the Chinese AI models. Essentially, Anthropic is saying that Chinese AI models are copying, doing something like a student might do during an exam when they peek at their friend’s answer sheet and copy the answers. In the AI world this process is called distillation, or in other words training smaller or new AI models on the output of a better AI model. Companies like Anthropic and OpenAI do this internally.
In a statement released this week, Anthropic alleged that Chinese companies created approximately 24,000 fraudulent accounts, which together made more than 16 million exchanges with the cloud. According to Anthropic, the goal was to extract high-quality outputs in areas such as coding, reasoning, and tool usage, strengths the company says define its model. “We have identified industrial-scale campaigns by three AI labs—DeepSeek, Moonshot, and MiniMax—to illegally extract the cloud’s capabilities to improve their own models,” Anthropic alleges.
While Anthropic acknowledges that distillation can be a legitimate method, it argues that it is illegal when used to copy powerful AI without permission. Anthropic says, “Models created through illegal distillation are unlikely to retain those safeguards, meaning dangerous potential could increase if many protections are removed entirely.”
Same same but different?
While Anthropic is making some serious allegations against Chinese models, the internet’s reaction has been sympathetic.
Social media quickly filled with memes stating that the entire AI industry is largely dependent on data scraping. Many users pointed to Anthropic’s own legal troubles, citing recent lawsuits where the company agreed to pay at least $1.5 billion to settle a lawsuit filed by authors who accused the company of using pirated books to train its cloud AI chatbot.
“I have a website about traditional Chinese medicine that I literally spent years building. When I asked Cloud a question about this topic, it was almost word-for-word exactly what I wrote myself. So please save us the gaslighting about training AI on the work of others,” one user wrote on X.
“This is completely evil. First, Anthropic used torrents to download books and data for training over the top of the internet, without anyone’s permission. Second, Chinese companies paid for the API and didn’t get the information for free. Third, it looks like Anthropic is tracking people,” another user claimed.
And when there’s controversy involving a competitor, tech billionaire Elon Musk rarely misses the opportunity to get involved. The XAI boss took direct aim at Anthropic and accused the company of conducting training on scraped data and paying for such practices. “Anthropic is guilty of stealing massive amounts of training data and has had to pay billions of dollars for the theft. That’s just a fact,” Musk wrote.
Of course, Musk’s own company, XAI, isn’t entirely immune to similar allegations. Responding to a user who pointed out that his company also trains on huge datasets, Musk acknowledged that AI systems are indeed trained on large amounts of data. However, he argued that the difference is that Anthropic is being “complacent” and “hypocritical” about it.
Attack on DeepSeek by the US government
Anthropic’s allegations also come at a time when the AI race between the US and China is intensifying, with the US government increasingly making advanced AI a matter of national security.
Further heating up the US-China AI standoff, a recent report from Reuters claimed that DeepSeek trained its latest AI models using Nvidia’s most advanced Blackwell chips despite US export restrictions. US officials reportedly believe the chips were clustered in a DeepSeek data center in Inner Mongolia, raising questions about whether export controls were bypassed.
Coincidentally, the attack by both the US government and Anthropic on Chinese AI companies comes at a time when DeepSeek is set to release its next AI model. The company, which entered the public conversation with the launch of the sensational R1 AI model, is expected to release its next big AI model within a week. There are rumors that when the AI model, expectedly named DeepSeek R2, is released, it will match or beat America’s best AI model.
The problem for American AI companies so far has been that Chinese AI models are mostly open-source models, with their training weights made available to everyone. This allows people and companies around the world to use them in ways they can’t use US AI models, which are proprietary. In such a scenario, if Chinese AI models are as good or nearly as good as Cloud, Gemini, and ChatGPT, it reduces the business value of American AI companies.
