The open-weight model is the unit of competition that Washington is now treating as a potential sanctions target. When a lab publishes its trained parameters publicly, those weights encode the distilled output of enormous compute investment. Any party with sufficient hardware can replicate or study that work, without paying for the training run that produced it. Treasury Secretary Scott Bessent said the United States could impose sanctions on China over what he called AI model "theft," a statement that ties punitive trade policy directly to this technical artifact.
The mechanism behind the warning
Chinese open-weight models have been gaining ground against leading American offerings from OpenAI and Anthropic. The competitive pressure follows directly from what the open-weight format permits: closed API models keep weights proprietary and accessible only through an interface, while open-weight releases hand downstream actors the raw parameters. Once published, those parameters cannot be recalled. The format that enables broad collaboration also makes appropriation harder to trace and harder to prove.
The allegation embedded in the word "theft" is that Chinese models have benefited from American research beyond what any open licensing terms would permit. Bessent's statement does not specify whether the conduct at issue involves pre-training data or actual weight derivation.
Where sanctions authority would fit
Using sanctions as a response to AI model appropriation would be a novel application of Treasury's enforcement toolkit. The Office of Foreign Assets Control targets financial flows, individuals, and entities. Extending that reach to cover intellectual property in AI model weights requires framing the conduct as a national security threat rather than a standard trade dispute.
Bessent's language does exactly that work. "Theft" places the complaint outside World Trade Organization remedy territory and into the national security register that has historically supported OFAC action. Whether that legal theory holds, and which agency would execute it, is not addressed in the statement.
What the commercial stakes look like for OpenAI and Anthropic
OpenAI and Anthropic are the American labs identified in the context of the competitive threat. Neither company's specific products or market position is characterized further in the statement. The pressure from Chinese open-weight alternatives is described as ongoing, and it continues regardless of what any sanctions process ultimately produces.