OpenAI Tells Congress: DeepSeek Is Ripping Off Our AI Models
OpenAI sent a memo to US lawmakers accusing Chinese rival DeepSeek of using distillation to freeload off American AI systems.
OpenAI is going to Washington with a pointed message: DeepSeek is cheating.
In a memo sent to US lawmakers, OpenAI accused its Chinese competitor of leveraging distillation techniques to train the next generation of its R1 model. The allegation? DeepSeek is essentially extracting knowledge from leading American AI models to bootstrap its own — a "free-ride" on US innovation, as OpenAI put it.
Distillation is a well-known ML technique where a smaller model learns to mimic a larger, more capable one. It's legal in many contexts, but OpenAI is framing DeepSeek's approach as unfair and increasingly sophisticated.
The memo is a clear play to get lawmakers on board with tighter restrictions around AI model access and exports. OpenAI wants regulators to see this as a national competitiveness issue, not just a corporate rivalry.
The AI cold war just got another chapter.