OpenAI told US lawmakers that Chinese AI company DeepSeek relied on US models for distillation training, effectively “free-riding” on outputs from OpenAI and other firms. According to OpenAI, DeepSeek employees used masked connections to bypass restrictions and extract model responses.
