Anthropic Flags 24,000 Fake Accounts in Claude Access Case
Synopsis
Anthropic has accused Chinese AI developers DeepSeek, Moonshot AI, and MiniMax of using fake accounts to extract outputs from its Claude language model, generating more than 16 million interactions. The company says this was done to train their own AI systems through distillation, a method of transferring capabilities. The claims arise as U.S. policymakers consider whether to tighten export controls on advanced AI chips.
Anthropic accused Chinese AI companies DeepSeek, Moonshot AI, and MiniMax of improperly using its Claude model via around 24,000 fake accounts to improve their own systems. The firms generated over 16 million exchanges using distillation, a training method that leverages outputs from stronger models, amid ongoing U.S. discussions on AI chip export restrictions.
Key Highlights
- Anthropic alleges three Chinese AI firms used Claude outputs to train their own models.
- Companies used about 24,000 fraudulent accounts and generated over 16 million interactions.
- Activity involved “distillation,” training weaker models from stronger output responses.
- Accusations come amid U.S. debate over export controls on advanced AI chips.
Anthropic claimed to have detected what it termed as mass abuse of its Claude system of artificial intelligence by numerous Chinese AI firms, claiming that they had utilised the results of the system to enhance their systems.
The San Francisco-based company called DeepSeek, Moonshot AI, and MiniMax made its claims. The firms had made about 24,000 fake/fraudulent accounts, according to Anthropic, to gain access to Claude and had created millions of interactions to elicit scale responses.
The process, according to Anthropic, was called distillation. Through distillation, a smaller model or a more recent model is being trained, learning by looking at the outputs produced by another more advanced model. The company claimed that the act contravened its terms of service and access.
Large-Scale Account Activity Detected
Anthropic reported finding odd usage behaviour, such as automated queries and coordinated account behaviour, which it associated with attempts to collect high amounts of Claude responses. The company claimed to block the accounts and added more monitoring and security measures.
The accusations reveal that the U.S. policymakers deliberated on tighter export controls on advanced semiconductors to train and operate artificial intelligence systems. Large language models are dependent on high-performance chips because they demand large amounts of computing resources and data processing capabilities.
AI Industry and Competitive Context
The worldwide AI industry has grown very fast, and huge funds are being pumped into the creation of models and cloud computing. Claude and other large language models are trained on large datasets, and they consume large amounts of hardware. Availability of both computing chips and high-quality training data has become a core issue of competition among developers of AI.
Anthropic, which was started in 2021, is in competition with other leading AI companies that provide enterprise and consumer AI services. Investors, such as Amazon, have been able to give the company billions of dollars in strategic support to build up and expand its AI systems.
Anthropic stated that it is cooperating with the authorities on the case. The mentioned companies have not described how people reacted to the alleged accusations.
Follow Inspirepreneur Magazine for the breaking news.
Pooja Malik is a business journalist with over six years of experience covering startups, entrepreneurship, and emerging trends. She has previously worked with leading media platforms such as YourStory Media and BW BusinessWorld, where she reported on business, policy, and market developments. Currently, she serves as Editor at The Inspirepreneur Magazine, where she writes and edits stories across business, lifestyle, and travel, with a focus on clarity, accuracy, and reader relevance.
