Anthropic has accused Alibaba of conducting the largest known "distillation attack" to date against a US artificial-intelligence company, telling US senators and White House officials that operators tied to Alibaba's Qwen AI lab used nearly 25,000 fraudulent accounts to extract the capabilities of its Claude models. The accusations, contained in a letter the company sent on June 10, 2026, were widely reported this week and mark the first time Anthropic has named a major Chinese technology conglomerate as the source of such a campaign.
According to Anthropic, the operation ran from April 22 to June 5, 2026, generating roughly 28.8 million interactions with Claude. The campaign specifically targeted Claude's most advanced and commercially valuable abilities: software engineering and agentic reasoning.
Key Highlights
- Anthropic alleges operators linked to Alibaba's Qwen lab used nearly 25,000 fraudulent accounts.
- The campaign generated roughly 28.8 million interactions with Claude between April 22 and June 5, 2026.
- It targeted Claude's most valuable skills: software engineering and agentic reasoning.
- This single campaign exceeded the combined volume of earlier alleged attacks tied to DeepSeek, MiniMax, and Moonshot AI.
- Alibaba's US-listed shares fell more than 3 percent on the news, dropping below 100 dollars.
What Is a Distillation Attack
Model distillation is the practice of feeding carefully constructed queries to a powerful frontier model, collecting its responses, and using those outputs to train a cheaper rival system that approximates the original's capabilities — for a fraction of the development cost.
When done against a competitor's commercial model in violation of its terms of service, Anthropic frames the technique as "adversarial distillation." The company warns that models built this way often inherit capabilities without the safety guardrails of the original, creating risks that go beyond intellectual property alone.
Details
The alleged Alibaba campaign was, by Anthropic's account, far larger than anything it had previously disclosed. In February 2026, the company said similar efforts linked to the Chinese startups DeepSeek, MiniMax, and Moonshot AI had collectively generated more than 16 million interactions through about 24,000 fake accounts. The single Alibaba operation alone is said to have surpassed that combined total.
Anthropic's letter noted that the campaign took place after a White House memo from policy adviser Michael Kratsios warning against such extraction, framing the activity as occurring in defiance of the administration's stated position.
The scale was made possible in part by a wider grey market for Claude access. According to unverified reports circulating on a technical forum, Chinese resellers have offered Claude capacity at 70 to 90 percent below official API prices, funded by pooling fraudulent accounts and reselling interaction data as training material to Chinese AI labs.
Impact
For developers and enterprises, the dispute sharpens a growing tension in the AI market: the line between legitimate model evaluation and the wholesale copying of a competitor's capabilities. If distillation at this scale becomes routine, frontier labs may tighten API access, add stricter account verification, and expand monitoring — changes that could ripple down to ordinary API customers.
Anthropic has urged the Trump administration to clarify antitrust guidelines so that US labs can share more information with one another about distillation attempts without running afoul of competition law. The threat has reportedly pushed Anthropic, OpenAI, and Alphabet toward greater coordination on detecting terms-of-service violations.
Background
The accusation lands amid escalating friction between Washington and Chinese technology firms. Alibaba was added to the Pentagon's list of Chinese military-linked companies and this week sued the Defense Department to challenge that designation. In the Senate, Republican Bill Hagerty and Democrat Andy Kim are planning an amendment to a pending defense package that would allow sanctions against Chinese firms that illicitly extract data from US AI models, with a related bipartisan effort in the House. It remains unclear whether the measures will gather enough support. Alibaba has not commented on the allegations.
What's Next
Expect frontier providers to harden their defenses against bulk extraction — tighter onboarding, anomaly detection on usage patterns, and clearer enforcement against reseller networks. For the MENA region, where data-sovereignty rules such as Tunisia's INPDP framework and Saudi Arabia's PDPL already shape how teams adopt foreign AI, the episode is a reminder to favor providers with transparent governance and to treat suspiciously cheap "grey-market" API access as a security and compliance risk, not a bargain.
Source: The Next Web