OpenAI has accused Chinese AI startup DeepSeek of using model distillation techniques to replicate its ChatGPT technology, according to a memo sent to U.S. lawmakers. The Sam Altman-led company claims DeepSeek employees are actively trying to circumvent access restrictions and use obfuscated third-party routers to harvest data from American AI models.

DeepSeek, backed by Chinese hedge fund High-Flyer, shook markets last year when its AI models demonstrated capabilities rivaling top U.S. offerings despite export controls. The Hangzhou-based company's rapid progress has fueled concerns in Washington about China's ability to close the AI gap. OpenAI alleges DeepSeek developed custom code to systematically access U.S. models and obtain outputs for distillation.

This revelation comes amid growing tensions in the global AI race, with Silicon Valley executives previously praising DeepSeek's DeepSeek-V3 and DeepSeek-R1 models. OpenAI says Chinese large language models are "actively cutting corners" on safety protocols during training and deployment. The company says it proactively removes users attempting to distill its models for rival development, highlighting the escalating competition between U.S. and Chinese AI firms.