Anthropic accuses Alibaba of illegally extracting AI model data

The AI company claims millions of interactions with its Claude models were used in an effort to replicate advanced capabilities, adding a new dimension to ongoing debates over AI competition, intellectual property and national security.

U.S.-based artificial intelligence firm Anthropic has alleged that a large-scale effort linked to Chinese technology company Alibaba sought to replicate capabilities of its Claude AI models through a technique commonly known as model distillation, according to reports citing company communications with U.S. lawmakers.

The allegations were detailed in a letter sent by Anthropic to members of the U.S. Senate Banking Committee ahead of discussions on artificial intelligence and global technological competitiveness. The company described the activity as the most extensive attempt it has identified to date involving the extraction of capabilities from its AI systems.

Model distillation refers to a process in which outputs generated by a sophisticated AI model are used to train another system, potentially allowing developers to reproduce certain behaviours and performance characteristics without direct access to the underlying technology.

Anthropic claims that the activity took place over several weeks between April and early June 2026. According to the company, the operation involved tens of millions of interactions with Claude through thousands of accounts that it believes were created to facilitate the effort.

Concerns over AI intellectual property

The company alleged that individuals connected to Alibaba and its AI research unit, Qwen, were behind the activity. At the time the allegations surfaced, Alibaba had not publicly commented on the claims.

Anthropic said the scale of the operation significantly exceeded previous incidents it had identified involving other AI developers. Earlier this year, the company raised similar concerns regarding several Chinese AI startups, alleging attempts to use Claude-generated outputs to enhance competing models.

The latest claims highlight growing concerns within the AI industry about safeguarding intellectual property as competition intensifies among companies racing to develop increasingly capable models. As AI systems become more advanced and expensive to build, developers are paying greater attention to preventing unauthorised replication of their technologies.

The dispute also underscores broader challenges associated with monitoring how AI outputs are used once they become accessible through public or commercial interfaces.

AI competition meets geopolitical tensions

Anthropic’s allegations arrive amid heightened scrutiny in Washington over the protection of advanced AI technologies. U.S. policymakers have increasingly expressed concerns about the transfer of sensitive AI capabilities to foreign entities, particularly in areas with potential strategic or national security implications.

The company has advocated for closer cooperation between government agencies and AI developers to identify and respond to potential threats involving advanced models and related technologies.

The issue emerges against a backdrop of escalating geopolitical competition in artificial intelligence between the United States and China. Governments on both sides are investing heavily in AI development while introducing new policies aimed at protecting technological advantages and regulating access to advanced systems.

While Anthropic's allegations remain unverified publicly, the dispute is likely to fuel ongoing debates about AI governance, intellectual property rights and the safeguards needed to protect cutting-edge technologies in an increasingly competitive global market.