Anthropic accuses Alibaba of obtaining 'illicit' access to Claude

Anthropic has accused Alibaba of obtaining what it described as illicit access to Claude, alleging a large-scale campaign to extract the model’s capabilities and urging U.S. lawmakers to tighten controls on advanced AI systems.

The claim centers on a June 10 letter Anthropic sent to the U.S. Senate Banking Committee. According to reporting published by the Financial Times on June 24 and followed by additional coverage on June 25, Anthropic said the activity represented the largest campaign it had seen aimed at extracting Claude’s behavior and performance.

Anthropic said the alleged activity took place between late April and early June 2026. In that window, the company said the accounts involved generated more than 28.8 million interactions with Claude.

The company also said the operation used about 25,000 fraudulent accounts. The reporting says Anthropic linked the effort to Alibaba and said the targeted capabilities included agentic reasoning, software engineering, and long-horizon tasks.

What Anthropic alleges

Anthropic’s accusation is not framed as a narrow abuse case. It is making a broader claim that a major Chinese technology company, or actors tied to it, tried to probe and extract frontier model behavior at industrial scale.

The company’s description matters because it goes beyond ordinary product misuse. By focusing on repeated interactions, account volume, and the specific capabilities targeted, Anthropic is suggesting an organized attempt to study how Claude responds across tasks rather than casual use.

Anthropic has previously accused other Chinese AI labs, including DeepSeek, Moonshot, and MiniMax, of distillation-style attacks on Claude. That background does not prove the Alibaba allegation, but it shows the company has been publicly treating model extraction as a recurring threat.

The timeline

The earliest date in the reporting is late April 2026, which Anthropic identified as the start of the alleged access campaign. The activity then continued into early June.

Anthropic’s letter to the Senate Banking Committee is dated June 10, 2026. The Financial Times reported on the allegation on June 24, and a follow-up FirstFT item repeated the same core figures later that day.

Additional coverage on June 25 repeated the allegation and said Alibaba declined to comment. At this point, the public record described in the reporting consists of Anthropic’s claim, the Senate letter date, and the subsequent media coverage.

Why Congress matters

Anthropic is not presenting this only as a commercial dispute. In its letter, the company urged Congress to close loopholes that it says allow PRC AI labs to access advanced U.S. technology.

It also called for penalties against PRC labs responsible for distillation attacks. That language places the allegation squarely in the center of U.S. policy debates about model protection, export controls, and how to limit access to frontier systems.

The Senate Banking Committee is relevant because the allegation is being presented to lawmakers rather than only to regulators, customers, or the public. That raises the chance of follow-up questions, hearings, or requests for the letter to be made public.

Alibaba’s response

The reporting cited so far does not include a substantive public rebuttal from Alibaba. The Financial Times said the company did not respond to a request for comment, and Investors.com later reported that Alibaba declined to comment.

That means the allegation remains unverified in the public record beyond Anthropic’s account and the surrounding reporting. There is no independent technical confirmation in the cited sources of the full evidence Anthropic used to attribute the activity to Alibaba-linked actors.

The lack of a direct response leaves the story in a developing phase. A denial, an explanation, or a technical counterargument from Alibaba could materially change how the allegation is understood.

Stakes for the AI industry

The dispute touches several overlapping risks: potential theft or extraction of frontier AI model capabilities, congressional pressure for tighter controls, and broader U.S.-China technology tensions.

For Anthropic, the allegation reinforces its argument that frontier models are now targets for systematic probing and extraction. The company is effectively warning that repeated access can reveal capabilities that should not be easy to copy or replicate.

For Alibaba, the claim introduces reputational and market risk at a sensitive moment. It also adds to the wider scrutiny faced by Chinese AI firms as U.S. policymakers debate how to protect advanced systems and intellectual property.

What to watch next

The next key developments are straightforward: whether Alibaba issues a direct denial or explanation, whether Anthropic releases more of the technical basis for its claim, and whether the Senate Banking Committee makes the June 10 letter public.

It will also matter whether lawmakers respond in a formal way, such as a hearing, a request for additional testimony, or renewed discussion of measures aimed at distillation attacks and access loopholes.

If independent security researchers or other outlets verify the alleged access pattern, that would strengthen the story materially. If they do not, the allegation may remain a significant but unconfirmed corporate accusation with policy consequences.

For now, the verified facts are limited but consequential: Anthropic says Alibaba-linked actors used tens of thousands of fraudulent accounts and tens of millions of Claude interactions in an alleged extraction campaign, and Anthropic has taken that claim directly to U.S. lawmakers.