US says Alibaba and DeepSeek siphoned AI models since 2024
In short
US agencies say Alibaba and DeepSeek used a method called distillation since 2024. They say it let the Chinese firms pull information from US AI models at an industrial scale to train their own systems.
This matters because it could affect AI competition and intellectual property disputes. The reporting does not say what punishment, if any, may follow.
What the reporting says
US agencies said Alibaba and DeepSeek have used distillation since 2024 to access and draw information from US AI models at industrial scale to train their own systems. The report says the technique was used to siphon model information, but it does not name any penalties or response actions. The exact models affected and the scope of the investigation are not stated in the available reporting.
Why it matters
US agencies said Alibaba and DeepSeek used distillation since 2024 to access and draw information from US AI models at industrial scale. The claim suggests a policy and national-security concern tied to AI model training and intellectual property, but the reporting does not state any enforcement action, penalty, or named affected model. The immediate consequence remains unclear.