Anthropic has declared war on what it describes as a massive, coordinated effort by Chinese tech companies to siphon knowledge from its flagship AI model, Claude. The company is now deploying identity verification, time-zone monitoring, and aggressive account purges to shut down access routes that have allegedly been exploited through overseas subsidiaries and proxy services.

The scale of the alleged extraction is staggering. According to the Financial Times, Anthropic accused Alibaba-affiliated entities in mid-June 2026 of orchestrating the largest extraction campaign it has ever detected, involving more than 25,000 fraudulent accounts that generated approximately 28.8 million exchanges with Claude.

What distillation campaigns look like

Anthropic had already uncovered a prior distillation campaign involving roughly 24,000 fake accounts and more than 16 million exchanges. The Alibaba-linked operation that surfaced in June was even larger, suggesting the problem was accelerating rather than shrinking.

How Anthropic is fighting back

Starting in April 2026, Anthropic implemented government ID verification through Persona, a third-party identity platform, to ensure that the humans behind API accounts are who they claim to be.

Anthropic has also deployed what amounts to informal surveillance techniques, monitoring account indicators such as time zones, query patterns, and usage volumes to flag suspicious activity.

The company has been explicit about its policy: commercial access to Claude is prohibited in China and for any entities controlled by the People’s Republic of China. This isn’t just a corporate preference. It’s a compliance posture aligned with US national security regulations and export controls governing advanced AI technology.

The bigger picture: US-China AI competition

The involvement of names like Ant Financial and Alibaba adds a layer of geopolitical sensitivity. These aren’t small startups trying to punch above their weight. They’re among China’s largest technology conglomerates, and allegations that their affiliates orchestrated extraction campaigns involving tens of thousands of fake accounts suggest a level of organizational sophistication that goes well beyond a few rogue developers.

What this means for investors

If Chinese firms are investing this heavily in extracting capabilities from Western AI models, it suggests that the gap between US and Chinese frontier models remains significant enough to justify industrial-scale espionage. That gap is what underpins much of the valuation premium currently assigned to leading US AI companies, including Anthropic, which was last valued at $61.5 billion.