OpenAI just sent a serious warning to Congress. The company claims that DeepSeek, a growing Chinese AI startup, is aggressively targeting American tech companies to copy their work and train its own systems.
In a memo sent Thursday to the House Select Committee regarding competition with China, Sam Altman’s team accused DeepSeek of trying to “free-ride” on the hard work done by OpenAI and other top U.S. labs.
The core of the complaint focuses on a method called “distillation.” Instead of training a model entirely from scratch—which costs billions of dollars and takes massive computing power—distillation uses a smarter, existing AI to teach a newer one. The established model grades the new model’s answers, effectively transferring its “brainpower” to the copycat. OpenAI says DeepSeek is doing exactly this with ChatGPT.
The memo gets specific about the tactics involved. OpenAI says it caught accounts linked to DeepSeek employees trying to sneak around security blocks. They allegedly used disguised third-party routers and wrote custom code to siphon data from U.S. models without revealing their true location.
This matters because DeepSeek is already a serious threat. The Hangzhou-based company shook up the financial markets early last year when it released AI models that rivaled the best American software. Those releases, DeepSeek-V3 and DeepSeek-R1, proved that China might be catching up in the AI race faster than Washington expected, despite U.S. trade restrictions. Even Silicon Valley executives have admitted that DeepSeek’s technology is impressive.
OpenAI argues that this speed comes with risks. They claim Chinese developers are “cutting corners” on safety to deploy these models as fast as possible. In the meantime, OpenAI says it is fighting back by banning any users it suspects of trying to steal its model’s data.
