OpenAI has alleged that Chinese AI startup DeepSeek recently tried to bypass its safeguards to obtain model outputs. According to Reuters, the company sent the memo to the US House Select Committee on Strategic Competition with China.
Sam Altman’s AI company claimed it identified accounts allegedly linked to DeepSeek employees who were attempting to work around access limits.
OpenAI also alleged that these users routed requests through third-party systems that masked their origin, which allowed programmatic extraction of responses at scale.
The dispute centers on distillation. The technique trains a smaller model using outputs from a stronger system. OpenAI alleged that DeepSeek used this method to build competing models without matching the original research effort. It described the process as ‘free-riding’ on US frontier labs.
The memo also stated that the company found code designed to automate access to American AI systems and collect outputs for training.
Also Read: Alibaba Enters Physical AI Race With New Robotics Model
OpenAI told lawmakers that some large Chinese language models made it easy to train and deploy safely. It framed the issue as a combination of commercial risk and national security concern. The submission comes as Washington debates tighter controls on advanced AI technologies.
DeepSeek drew global attention last year after releasing high-performing AI models at a lower cost. The breakthrough raised fears that China could close the AI gap despite export restrictions.
The latest allegations are likely to intensify calls for stronger model protections and clearer global rules for AI development.