OpenAI has informed U.S. lawmakers that DeepSeek employed distillation and other evasive methods to extract output from U.S. AI models in order to train its R1 chatbot, which has raised concerns about business and national security. This revelation aligns with heightened scrutiny from lawmakers regarding foreign AI entities using similar techniques, such as proxy rotation and query randomization, to systematically scrape API outputs while evading usage policies.
OpenAI raises national security concerns about DeepSeek’s use of distillation methods
