White House Scrutinizes Chinese AI App DeepSeek Over Security and IP Risks
The White House and U.S. intelligence officials are examining whether the Chinese AI application DeepSeek poses national security risks and whether it may have involved intellectual property theft. OpenAI reportedly found evidence that DeepSeek used its AI models for training through a technique called distillation, which the company says violates its terms of service.
What happened
OpenAI said it found evidence that Chinese AI startup DeepSeek used OpenAI's AI models to train its own open-source competitor through a technique called distillation. The company stated that distillation, where a model learns from a larger, more capable model to achieve similar performance at lower cost, violates OpenAI's terms of service when used to create competing models.
The White House National Security Council is analyzing the national security implications of the DeepSeek application. White House press secretary Karoline Leavitt described it as 'a wake-up call for the American AI industry.'
Reported impact
David Sacks, the White House AI and cryptocurrency czar, stated there is 'substantial evidence' that DeepSeek distilled knowledge from OpenAI's AI models and compared it to theft, though he did not specify what evidence he possessed.
Index analysis
Distillation is a common practice in the industry where developers use a larger, more performant model to calibrate a smaller model, reducing costs and latency. OpenAI's terms of service prohibit modifying or copying any service, extracting data and results, or using results to develop models that compete with OpenAI.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.