Not just Gemini, someone is trying to clone Anthropic’s Claude as well

The global competition in artificial intelligence is entering a new, tense chapter that pits American companies against Asian competitors.

Anthropic has recently launched a strong accusation against three well-known Chinese companies, namely DeepSeek, Moonshot and MiniMax, accusing them of orchestrating industrial-scale campaigns to illicitly appropriate the technologies underlying their systems.

According to the released data, the operation would have involved the creation of around 24,000 fake accounts, purposely designed to circumvent geographic restrictions and violate the terms of service. Through these fraudulent profiles, the involved entities would have generated over 16 million interactions with the Claude assistant.

The American company did not hesitate to frame the matter as a real national security issue, raising concerns that go far beyond simple commercial competition.

AI under attack, according to Anthropic the blame lies with the Chinese

The accusations focus on a specific technique defined as a distillation attack. In computing, distillation represents a completely legitimate and widely adopted method by labs to create lighter and cheaper versions of their software, intended for end users.

However, the same technique can be used maliciously to perform reverse engineering. By repeatedly querying an advanced system with countless variations of prompts, a rival company can acquire and copy complex logical capabilities in a short time.

Moreover, this shortcut allows cost reductions compared to independently developing native technology.

Considering that major US companies are spending tens of billions of dollars on infrastructure, research and data centers, the possibility that foreign actors could replicate such architectures at low cost represents an economic threat of colossal proportions, capable of erasing any competitive advantage.

Is intellectual property theft now a problem for AI companies?

The complaint fits into a regulatory and ethical landscape that is decidedly complex and full of contradictions. Already in January OpenAI had pointed the finger at DeepSeek for substantially identical reasons.

Yet, the complaints of Western tech companies often evoke ironic reactions from the public and professionals. Most of these same companies have in fact trained their algorithms on huge volumes of copyrighted works, without seeking permissions or acknowledging financial compensation to the rightful owners.

The recurring justification is that a strict approach to copyright would hinder the development of cutting-edge systems, advantaging nations that do not apply the same limitations.

In this regard, during a sector event held in July 2025, US President Donald Trump had argued that machine learning should not be considered a copyright violation, drawing a parallel with human knowledge acquisition through reading and noting how Chinese rivals openly ignore such constraints.

This positions the US industry in a rather ambiguous situation, committed to inflexibly defending its intellectual property on final models, but decidedly more permissive when it comes to acquiring the data needed to build them.

A global call to action

Despite clear violations of the platform’s terms of use, it remains uncertain whether the extraction actions carried out by the three Chinese labs directly infringe current international laws, or if there are effective legal tools to curb them definitively, beyond the temporary suspension of the incriminated accounts.

In the face of intrusions that are becoming every day more elaborate, like the one also highlighted by Google on its Gemini chatbot, the leadership of Anthropic has stressed that the window of time to intervene is rapidly shrinking.

The company strongly calls for swift and coordinated action involving industry players, political institutions and the global community, in order to curb a phenomenon that threatens to compromise Western technological future.