STORY · FORSKNING_
Research: Censorship from DeepSeek is not transferred when distilling into GPT-OSS
Researchers distilled a U.S. AI model (GPT-OSS-120B) based on outputs from DeepSeek V4 Flash to improve financial reasoning. Despite the teacher model showing political censorship on China-sensitive questions, the student model showed no corresponding censorship behavior and answered freely to the same questions.
WHY IT MATTERS
The finding addresses concerns in Washington and regulatory industries that foreign influence and values are transferred with open AI technology from China. The study suggests that desired domain knowledge transfer can be achieved without inheriting political censorship.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.