US Report Details How China Is 'Ripping Off' Frontier AI From Anthropic and OpenAI

A new US-focused report walks through evidence that Chinese labs are systematically probing, distilling, and reverse-engineering outputs from Anthropic's Claude and OpenAI's GPT models to accelerate their own frontier work. Whether or not you accept the report's tone, its findings will shape the next round of US export-control policy.

What the report claims

Related: GPT-5 Is Here: Everything You Need to Know About OpenAI's Most Powerful Model Yet →

The techniques described include high-volume API querying to build training corpora, targeted prompt strategies to extract reasoning chains, and coordinated use of open Chinese frontier releases (Qwen, DeepSeek) as scaffolding around distilled American capabilities. None of this is new individually; the report's contribution is documenting the scale.

The policy fallout

Related: Will AI Coding Agents Replace Developers? We Asked 100 Engineers →

Expect Congress to use the report to push tighter API access controls for entities on the Commerce Department's watchlists, and to accelerate proposals for know-your-customer requirements on frontier API access. US labs already do some of this quietly; formalizing it changes contract law, not just security posture.

The counterargument

Related: The 27 Best AI Tools in 2026 (Tested for 90 Days) →

Distillation from public API outputs is inherently hard to prevent without breaking legitimate research access. Chinese labs are also producing genuinely novel work — DeepSeek's efficient architectures were not stolen. The risk of overreacting is a bifurcated global AI ecosystem in which US labs lose research reach as much as adversaries do.

The bottom line

Related: ChatGPT vs Claude 4: Which AI Should You Actually Pay For in 2026? →

This report will become a data point in every US AI policy conversation for the rest of the year. Expect real, consequential changes to how American frontier labs sell API access abroad.