OpenAI accuses DeepSeek of using data from American models to train R1

# OpenAI Accuses DeepSeek of Unauthorized Use of American Models

Tensions in the global artificial intelligence race have reached a new level. OpenAI has issued an official warning to American lawmakers that Chinese startup DeepSeek employs sophisticated and cunning methods to extract valuable information from leading American neural networks with the goal of training its breakthrough chatbot R1. This incident, reported by Bloomberg, reveals not just a commercial conflict, but a fundamental confrontation over intellectual property protection and competitive advantage in the age of artificial intelligence.

According to an internal OpenAI memo, DeepSeek uses a distillation technique — a process in which a smaller model learns from the outputs of a larger and more advanced system. In practice, this means that DeepSeek engineers systematically submit queries to American models, analyze their responses, and use the obtained data to train their own algorithms. This approach allows the Chinese company to save enormous computational resources and development time, which are typically required to create models of comparable performance levels. In other words, according to OpenAI, DeepSeek is essentially parasitizing on the investments of American companies in research and development.

The context of this conflict is critically important for understanding its scale. Developing advanced language models requires hundreds of millions of dollars for purchasing computing power, hiring top specialists, and conducting experiments. OpenAI, Microsoft, and other American giants have spent billions to reach the current level of performance. DeepSeek, by contrast, managed to reach a comparable level in much shorter timeframes and with significantly lower costs. This raises legitimate questions about whether this success was achieved honestly or is the result of systematic "borrowing" of knowledge from American models.

OpenAI's accusation is important also because it focuses attention on the vulnerability of API interfaces. Most advanced models are available through web services, where developers can submit requests and receive responses. Although terms of use prohibit using output data to train competing systems, this is extremely difficult to track and prevent. DeepSeek potentially simply executed this scheme at scale, processing millions of requests and extracting structured information from them for its needs.

The incident also highlights the growing imbalance in the technological race between the US and China. American companies face strict restrictions on exporting advanced technology to China, but they cannot completely block access to their public services without harming global business. DeepSeek apparently found a loophole in this system. Now OpenAI is calling for strengthened intellectual property protection and more rigorous monitoring of commercial API usage.

The further development of events depends on the response of American regulators and the measures they take to protect local industry interests. In the short term, this could lead to stricter conditions for accessing advanced models and the implementation of more complex verification systems. In the long term, the conflict underscores the need to reconsider approaches to sharing intellectual property in the age of open APIs and the global digital economy. The question of how to protect innovation and investment while remaining open to scientific progress remains one of the most pressing in the industry.