An unverified claim circulating on social media alleges the US government has detected watermarks from American AI models embedded within Chinese AI systems. No official US agency has confirmed this assertion, and no major news outlet has independently verified it.
AI watermarking involves embedding invisible signatures into a model's outputs, weights, or training methodology. These marks are designed to persist even if the model is fine-tuned or distilled into a new version.
If true, the claim would imply Chinese AI models were derived from or trained using proprietary US technology. This comes amid escalating US export restrictions on advanced semiconductors and AI technology to China, targeting high-end chips and manufacturing equipment.
The US-China AI rivalry is intensifying, with companies like Baidu, Alibaba, and DeepSeek releasing models that benchmark close to leading American systems. Despite the claims, no crypto market reaction has been reported.