Why this score US Treasury Secretary. Speaks with the authority of office, but this cycle he asserts a technical forensic finding (model watermarks) on social media with no published evidence, as a principal in an active trade dispute — authoritative on policy intent, motivated and unverified on technical fact.
Tracked statements
2
? unverified
The United States remains the leader in AI. And we are open to discussions on avoiding shared risks and avoiding bifurcation of our two systems. We expect our discussions to cover both open and closed weight models.
A written statement to Axios, eight days after three US agencies accused six Chinese companies of industrial-scale distillation. Nine days later the two governments announced a standing AI dialogue; neither government’s account says whether it covers open- and closed-weight models.
“We are finding watermarks of our US large language models on many Chinese models, and that’s unacceptable.”
Unverified: no supporting evidence was published — neither Treasury nor Commerce released watermark findings, sample models, methodology, or named a company, and the July UK AISI/CAISI Kimi K3 assessment tested cyber capability, not provenance. The threatened consequences remain unexecuted: no Entity List addition or sanctions against any Chinese AI firm, and Moonshot is absent from the Pentagon's June Section 1260H list. This is an absence in the published record, not proof no evidence exists.