Chinese AI Startup DeepSeek Faces Scrutiny Over Alleged ChatGPT-Based Training

Microsoft and OpenAI have initiated an internal investigation over suspicions that Chinese AI developer DeepSeek illegally accessed data from U.S.-based AI systems. The incident, brought to light by anonymous Bloomberg sources, threatens to upend the competitive landscape of the global artificial intelligence race.

In fall 2023, Microsoft engineers detected unusual activity: massive data requests were being made via OpenAI’s API—a tool designed to integrate AI models into third-party applications. Traces of the activity, according to Microsoft, linked back to DeepSeek, whose recently unveiled R1 model stunned the market by claiming superiority over OpenAI and Meta’s systems in solving mathematical problems and data analysis. Investigators suspect DeepSeek employed “distillation”—a method of training its own AI on outputs from rival models—to drastically cut development costs.

The release of this “thinking model” triggered a stock market collapse among U.S. tech giants, with combined losses for Microsoft, NVIDIA, Oracle, and Alphabet exceeding $950 billion in a single day. The plunge followed DeepSeek’s announcement of a ChatGPT competitor purportedly ten times cheaper to operate. However, experts question the legitimacy of these claims. OpenAI has publicly accused Chinese firms of “systematically attempting to replicate their technology,” prompting tighter collaboration with U.S. authorities to safeguard intellectual property.