Story perspectives

Fraudulent Accounts Target U.S. AI Models Amid Export Restrictions

4/7/2026

Summary

Story summary

Anthropic reported that DeepSeek, Moonshot AI, and MiniMax conducted 16 million interactions with its Claude model via fraudulent accounts to bypass U.S. export restrictions.
The incident underscores a trend of industrial-scale distillation that threatens AI infrastructure as China tightens intellectual property policies.
The U.S. AI sector faces challenges securing models against unauthorized replication, with a potential two-tiered market that could undermine the value of regulated AI systems.

1 of 1

Story summary

Anthropic reported that DeepSeek, Moonshot AI, and MiniMax conducted 16 million interactions with its Claude model via fraudulent accounts to bypass U.S. export restrictions.
The incident underscores a trend of industrial-scale distillation that threatens AI infrastructure as China tightens intellectual property policies.
The U.S. AI sector faces challenges securing models against unauthorized replication, with a potential two-tiered market that could undermine the value of regulated AI systems.