Sep 30, 2026 DISPATCH // HARDWARE, CODE & PLATFORMS

Jensen Huang pushes back on AI theft claims as US warns of sanctions over model distillation

Nvidia's Jensen Huang rejects US accusations of AI model theft, calling model distillation fair competition as the US and China clash over AI technology.
Jensen Huang pushes back on AI theft claims as US warns of sanctions over model distillation Nerds Magazine © nerdsmagazine.com
Jensen Huang pushes back on AI theft claims as US warns of sanctions over model distillation © nerdsmagazine.com

US officials have called AI model distillation theft. Jensen Huang, Nvidia's CEO, doesn't flinch. He calls it competition. The fight over what counts as innovation and intellectual property is now front and center in the US-China tech standoff.

Huang's response goes straight at the White House and Treasury Secretary Scott Bessent. In July, Bessent called distillation "theft" and warned that sanctions could hit foreign companies copying US-built models. Huang didn't echo the outrage. On CNBC's Squawk Box, he said, "It's called competition." He says it's normal for tech companies to test and break down rival products. Nvidia's own products, he admits, get stripped "down to the bones" by competitors looking for an edge.

On September 8, 2026, the NSA, CISA, and FBI jointly accused China-based AI firms of conducting 'industrial-scale' distillation of U.S. frontier models since at least late 2024.

Reuters

The debate isn't just about ideas. Earlier this month, the US Cybersecurity and Infrastructure Security Agency accused Chinese AI firms of running massive knowledge-distillation campaigns. These campaigns, officials say, broke US companies' terms of use. Anthropic, a major AI company, also accused Alibaba's Qwen models and DeepSeek of illicit distillation. The US government's message is clear: unauthorized distillation is theft. Sanctions are on the table.

Reuters-linked reports say the NSA, CISA, and FBI advisory described "billions of tokens" and "millions of requests" aimed at advanced models like Claude, GPT, Gemini, and Grok. This scale has US officials worried about the security and ownership of top AI models. The advisory has become a key reference in the dispute, showing just how high the stakes are.

Huang draws a line. He says learning from rivals is part of the game, but criminal acts are not. In a perfect world, he says, no one could reverse-engineer a product. But he argues that competition drives progress. His advice to companies afraid of being copied is blunt. If you don't want others using your product, turn it off.

Anthropic reported earlier this month that it detected alleged illicit distillation involving Alibaba's Qwen models and DeepSeek, describing these incidents as part of a broader pattern of copying capabilities from leading frontier models.

CNBC

Two worldviews are colliding. One side wants protection. The other wants open competition. This is the new front in the US-China AI arms race. Accusations are flying. Sanctions could hit soon. The big question: is model distillation a crime or just the cost of progress? Huang doesn't back down. For the world's top AI companies, the fight isn't just in code. It's about the rules themselves. The US may threaten penalties. But as long as the fastest learners win, the real battle stays out in the open.

As of September 2026, the US government draws a line between legal distillation in research and unauthorized extraction from frontier models. The latter could mean sanctions. But as of September 28, 2026, no sanctions have been confirmed. The debate is still raging in public and in the industry. Huang's recent CNBC interview excerpts show just how unsettled the issue remains.

Topics:
Artificial Intelligence Tech Analysis #AI Models & Concepts #Model Distillation
Evan Solberg Technology publisher and editor-in-chief Nerds Magazine
Editor-in-Chief

Evan Solberg

Evan Solberg is the Founder, Owner, Publisher, and Editor-in-Chief of NerdsMagazine, where he covers consumer technology, software, artificial intelligence, privacy, and digital products. His editorial approach focuses on what technology actually does for readers, what it costs, where it falls short, and which claims deserve closer scrutiny.