Wire · founder news, decoded · regulatory
How China is ripping off cutting-edge AI from Anthropic, OpenAI
Anthropic and OpenAI have accused Chinese AI firms (Alibaba, DeepSeek, Minimax, Moonshot) of systematically stealing their models via distillation—training cheaper models on outputs from advanced Western AI without permission—to compress what would be an 18+ month technology gap into 6-9 months. The Trump administration and Congress are treating industrial-scale AI distillation as a national security threat comparable to chip export controls.
This Wire brief sits within Fusion42's coverage of AI Frontier Models. Wire is Fusion42's founder-focused intelligence feed: each story is connected to the funds and startups it names — every one with a live profile on Raise or Scout — so founders can follow the capital and the momentum behind the headline rather than just the headline itself. Wire analysis is one of the live surfaces Arthur, Fusion42's AI co-founder, reasons over.
The Wire takeaway
If you're building an AI model or service in the US, the government is about to treat your outputs as national security exports—expect API access restrictions, audit requirements, and watermarking mandates to become table stakes within months. The 18-month lead China loses without distillation is the exact window the Trump administration is trying to buy with regulation.
Read the full story at nypost.com →
Topics: AI Frontier Models · china-ip-theft · ai-distillation · export-controls · national-security · regulatory-shift