Anthropic’s Accusation Against Alibaba: A Significant Allegation
Anthropic has raised serious concerns in a confidential letter to the US Senate, accusing Alibaba of executing a massive distillation attack on its AI model, Claude. This alleged attack is described as the largest ever recorded, with indications that Alibaba created approximately 25,000 fake accounts to illegitimately extract capabilities from Claude without investing in its own development.
The Distillation Attack Explained
According to reports, the letter was sent around June 10, 2026, addressing notable senators Tim Scott and Elizabeth Warren. The letter details how operators associated with Alibaba and its AI laboratory, Alibaba Qwen, allegedly generated over 28.8 million interactions with Claude between April 22 and June 5. To avoid detection, these actors reportedly utilized obfuscation techniques and proxy networks.
The main goal of these attacks was to extract Claude’s advanced capabilities, particularly in fields such as agentic reasoning and software engineering. Given that Claude is unavailable in China, these fraudulent accounts were necessary to facilitate such unauthorized access.
Understanding AI Distillation
Distillation is a technique often employed in AI to train smaller and more cost-effective models using outputs from a more powerful model. Though not illegal in itself, large-scale unauthorized distillation operations, as alleged in this incident, raise ethical and legal concerns. Anthropic claims these actions could undermine billions of dollars of American investment, transforming American innovation into a financial boon for geopolitical competitors.
The Context of Previous Accusations
This is not the first time Anthropic has voiced concerns over unauthorized access by Chinese entities. Earlier in February, the company identified three Chinese labs—DeepSeek, Moonshot, and MiniMax—that had generated over 16 million interactions with their models using around 24,000 fraudulent accounts. However, the scale of the alleged Alibaba attack is notably larger, raising the stakes significantly.
Responding to Allegations
There is a notable sense of audacity in Anthropic’s accusations, especially considering Alibaba’s prominent position on the New York Stock Exchange and its operations in the United States. The timing of the alleged attacks, occurring after a previous warning from the Trump administration regarding the inadmissibility of such conduct, adds to the gravity of the issue.
The Financial Times highlighted that the White House had identified Alibaba’s associations with Chinese military operations against US interests—a claim the company fiercely denies.
Proposed Legislative Measures
In response to these threats, Anthropic is advocating for three specific legislative measures:
- Updating antitrust laws to allow sharing of information regarding these kinds of tactics among AI companies without facing legal repercussions.
- Strengthening export controls on advanced chips to prevent China from exploiting American AI models, even if they gain access through illicit means.
- Introducing legislation that would impose penalties on Chinese labs found to be undertaking distillation attacks, potentially blocking their access to American-made AI chips or models.
The Broader Implications
Anthropic warns that if China succeeds in bridging the technology gap, it could enhance its cyber capabilities, posing a serious threat to both the US government and American corporations. Anthropic’s Mythos model is highlighted as particularly concerning, already regarded as a “cyber nuclear weapon” by industry experts.
Future Prospects
The fallout from these allegations has been immediate, with Alibaba’s shares dropping over 4% in the Hong Kong market following the accusations. As this situation unfolds, the crucial question remains whether Anthropic can persuade Congress to support its legislative requests. For now, the company has pledged to work constructively with Congress and the Administration to safeguard American leadership in AI technology.
The outcome of this complex situation holds significant implications for the AI landscape, global technological competition, and the response of regulatory bodies to emerging threats in the digital age.