MASSIVE Chinese AI Theft Operation EXPOSED

Surveillance camera in front of the Chinese flag

American AI firm Anthropic has exposed a massive Chinese espionage operation involving 24,000 fraudulent accounts and 16 million illicit exchanges designed to steal U.S. artificial intelligence technology—a brazen act of intellectual theft that threatens national security and undermines billions in American innovation.

Story Snapshot

Anthropic publicly accused three Chinese AI firms—DeepSeek, Moonshot, and MiniMax—of using over 24,000 fake accounts to extract 16 million exchanges from its Claude AI model
The “distillation” operation violates U.S. export controls and terms of service, allowing Chinese companies to replicate advanced AI capabilities at minimal cost
Anthropic warned that stolen technology could feed into military, intelligence, and surveillance systems with safeguards removed
The accusations follow similar complaints from OpenAI and Google about Chinese AI theft, escalating the U.S.-China technology war

Chinese Firms Exploit American Innovation Through Massive Fraud Network

Anthropic revealed Monday that three Chinese AI companies orchestrated an industrial-scale theft operation targeting its Claude AI model. DeepSeek conducted over 150,000 exchanges focused on reasoning and censorship circumvention, while Moonshot AI executed 3.4 million exchanges targeting coding and agentic tasks. MiniMax ran the largest campaign with 13 million exchanges, pivoting its attack within 24 hours of each new Claude release. The companies deployed sophisticated “hydra cluster” networks mixing fraudulent extraction with legitimate-looking traffic to avoid detection, demonstrating state-level coordination and resources.

Export Control Violations Pose National Security Risk

The distillation technique allows Chinese firms to train smaller AI models using outputs from advanced American systems, effectively reverse-engineering technology protected by U.S. export controls. Anthropic emphasized that foreign labs conducting illicit distillation can strip away safety guardrails embedded in American models, potentially feeding unconstrained AI capabilities into Chinese military and surveillance apparatus. This represents exactly the scenario export restrictions aimed to prevent—adversarial nations obtaining cutting-edge AI technology while circumventing both legal constraints and the billions of dollars required for legitimate development.

Pattern of Chinese Technology Theft Escalates

Anthropic’s accusations build on recent complaints from other major U.S. AI companies. Earlier this month, OpenAI warned lawmakers that DeepSeek was distilling ChatGPT outputs, while Google reported agents querying its Gemini model over 100,000 times. Unlike previous vague allegations, Anthropic provided granular evidence including specific prompt patterns, account fingerprints, and real-time detection of ongoing campaigns. The timing proves particularly concerning as DeepSeek prepares to release its V4 model, likely leveraging stolen American intellectual property while Chinese firms face hardware limitations from U.S. sanctions.

American Companies Deploy Countermeasures and Call for Government Action

Following detection of the fraudulent networks, Anthropic deployed advanced classifiers and behavioral fingerprinting systems while tightening account verification procedures. The company shared threat indicators with other AI labs, cloud providers, and government authorities to enable coordinated defense. Anthropic explicitly called for industry-wide and policy responses as the campaigns grow increasingly sophisticated. This represents a critical test for the Trump administration’s commitment to protecting American technological superiority—Chinese state-backed entities are actively stealing innovations funded by U.S. investment, undermining our competitive advantage and potentially arming authoritarian surveillance systems.

Anthropic Accuses DeepSeek, Other China-Based AI Firms Of Free-Riding https://t.co/hieO9XW0VF

— zerohedge (@zerohedge) February 25, 2026

The exposure of this massive theft operation underscores why strong export controls and aggressive enforcement matter for national security. While critics may claim hypocrisy given U.S. companies’ data collection practices, there exists a fundamental difference between commercial activity and foreign adversaries stealing restricted military-relevant technology through fraudulent means. As artificial intelligence becomes central to economic and military power, allowing China to free-ride on American innovation while deploying the technology for authoritarian purposes represents an unacceptable threat that demands immediate government and industry action.