Sam Altman’s OpenAI accuses Chinese DeepSeek of copying AI models; writes letter to US lawmakers

Sam Altman-led OpenAI has accused Chinese DeepSeek of free-riding on the capabilities developed by OpenAI and other US frontier to replicate models and use them for its own training. The ChatGPT-maker has written a letter to the US House Select Committee warning them of DeepSeek using unfair and increasingly sophisticated methods to target OpenAI and America’s other leading AI companies. “Ahead of DeepSeek’s expected Lunar New Year release of a new, more powerful model, we are providing the Committee with an updated assessment of its evolving distillation tactics,” OpenAI writes in the memo. “OpenAI believes that infrastructure is destiny: chip development, power generation, transmission, and data center capacity will determine which countries can train and deploy frontier systems. This is why we’re investing through our Stargate Project to expand US AI infrastructure to 10 GW by 2029, and just one year in, we’re already over halfway toward that Goal,” the memo stated. “But we are equally focused on ensuring a level playing field, one where the People’s Republic of China (PRC) can’t advance autocratic AI by appropriating and repackaging American innovation”.Distillation tactic that OpenAI is accusing Chinese DeepSeek of involves using an older and more established AI model to review and judge the responses produced by a newer model. Through this process, knowledge and patterns learned by the older system are passed on to the newer one, helping it improve the quality of its answers.“We have observed usage patterns from several major Chinese LLM providers and some university research lab usage that are consistent with, and would be highly beneficial for, creating competitor models through Distillation,” OpenAI says in the memo.In the memo, OpenAI says that it has observed accounts associated with DeepSeek employees developing methods to circumvent OpenAI’s access restrictions and access models through obfuscated third-party routers and other ways that mask their source. “We also know that DeepSeek employees developed code to access US AI models and obtain outputs for distillation in programmatic ways. We believe that DeepSeek also uses third-party routers to access frontier models from other US labs,” it added.OpenAI says that it believes the best defense is offense: “the best way to ward off a fast-oncoming PRC making headway around the world for autocratic AI is continued investment in American AI leadership and global adoption of responsibly developed, democratic AI.” “We continue to lead in responsible frontier model innovation, invest across the full AI stack to train and deploy our systems safely, and make powerful AI tools available for free. Adoption of our latest agentic coding model, GPT-5.3-Codex, is up 60% week-over-week. Adoption of ChatGPT is at ~10% monthly growth,” the company added.