Alibaba Accused of AI Capability Extraction Campaign by Anthropic

AI research firm Anthropic has formally accused Chinese tech giant Alibaba of engaging in a significant and “illicit” effort to steal its advanced artificial intelligence capabilities. The accusation, detailed in a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs, marks a serious escalation in the ongoing global race for AI dominance and highlights the growing concerns surrounding intellectual property theft in the rapidly evolving AI landscape.

The letter, addressed to Senators Tim Scott and Elizabeth Warren on June 10, specifically details what Anthropic describes as the “largest known distillation attack on Anthropic to date.” Distillation, in the context of AI, is a sophisticated training technique where a smaller, less resource-intensive AI model is developed by leveraging the outputs of a much larger, more powerful foundational model. This process allows for the creation of specialized AI applications that can mimic the performance of their larger counterparts, often without the extensive computational resources or proprietary data required for initial training.

According to the letter, which CNBC has reviewed, operatives linked to Alibaba and its AI research division initiated a massive operation. Between April 22 and June 5, these entities reportedly conducted an staggering 28.8 million interactions with Anthropic’s AI models. This extensive engagement was allegedly carried out using approximately 25,000 fraudulent accounts, suggesting a deliberate and systematic attempt to extract valuable AI intelligence.

“We firmly believe that combating the threat of illicit distillation necessitates coordinated action between government and industry,” an Anthropic spokesperson stated. “We are committed to continued collaboration with Congress and the Administration to preserve American leadership in artificial intelligence.” This statement underscores Anthropic’s view that the challenge extends beyond a single company and requires a strategic, multi-stakeholder approach.

A representative for Alibaba did not immediately respond to requests for comment. The initial report on the letter was published by Bloomberg.

This recent accusation follows closely on the heels of a significant policy directive from the White House Office of Science and Technology Policy (OSTP) in April. The OSTP issued a memorandum pledging government support to help AI companies identify and counter industrial-scale distillation attacks. Anthropic’s letter pointedly notes that Alibaba’s alleged actions represent a disregard for these explicit warnings from the U.S. administration.

This is not the first time Anthropic has raised alarms about sophisticated AI theft. In February, the company publicly announced the detection of three separate “industrial-scale” distillation campaigns originating from other AI labs: DeepSeek, Moonshot, and MiniMax. At the time, Anthropic highlighted the increasing sophistication and intensity of these campaigns, calling for greater industry-wide collaboration among AI developers, cloud providers, and policymakers to establish robust defenses.

However, Anthropic’s engagement with policymakers has faced recent complications. Earlier this month, the company disclosed that it had received an export control directive from the Trump administration. This directive mandated Anthropic to immediately suspend access to its latest Claude models, specifically Fable 5 and Mythos 5, for “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” While the government cited national security concerns, it did not provide specific details regarding the nature of these concerns, according to Anthropic.

In response to this directive, senior Anthropic staff traveled to Washington D.C. for urgent meetings with members of the Trump administration. The company expressed optimism that both parties were working expeditiously to resolve the situation, but as of this report, there has been no definitive timeline provided for when access to these advanced models will be restored.

The implications of Anthropic’s allegations against Alibaba are far-reaching. Beyond the immediate concern of intellectual property theft, the incident raises critical questions about the security of AI infrastructure, the effectiveness of current regulatory frameworks, and the geopolitical ramifications of the global AI arms race. As AI technology continues its rapid ascent, the ability to protect proprietary models and prevent their illicit replication will be paramount to maintaining competitive advantage and national security.