OpenAI accuses China's DeepSeek of using its models for training
OpenAI says it has evidence that Chinese AI startup DeepSeek used its proprietary models to train its chatbot, as the developer of ChatGPT told the Financial Times.
- The American company found evidence of so-called "distillation," a technique that developers use to get better performance on simpler models by using the results of larger and more powerful models. This allows them to achieve similar results on specific tasks at a much lower cost.
- OpenAI declined to provide further details about its allegations.
- The company's terms of service state that users cannot copy its services or use the results to build models that compete with OpenAI.
- The publication notes that distillation is a common practice in the industry, and it is also common for AI labs in both China and the US to use the results of leading companies like OpenAI.
Did DeepSeek really steal OpenAI's intellectual propery?
According to Bloomberg, security researchers at Microsoft, which is OpenAI's largest investor, noticed that in the fall of 2024, people they believed to be connected to DeepSeek were uploading data en masse through the OpenAI application programming interface (API). It is through this API that software developers and business customers access OpenAI's services.
Microsoft has notified OpenAI of suspicious activity, and both companies are currently investigating the possible unauthorized acquisition of data by a group associated with DeepSeek.
David Sachs, White House advisor on artificial intelligence and cryptocurrencies, said in an interview with Fox News that there is a possibility that DeepSeek may have stolen US intellectual property.
When asked for comment, a representative of OpenAI said that Chinese companies and other entities are constantly trying to reproduce the models of leading American AI companies. At the same time, he stressed the need for close cooperation with the US government to protect advanced developments.
Earlier, former Meta engineer Yanshun Tai made an interesting observation in his post on LinkedIn: DeepSeek uses the OpenAI client library for API requests.
According to him, DeepSeek was able to save weeks of work on client libraries for Node.js and Python by using OpenAI's out-of-the-box solutions. In addition, developers can easily test or migrate to DeepSeek by changing only the base URL and API key in their code.
DeepSeek, which positions itself as an affordable alternative to American AI technologies, made waves in the market when its free AI assistant outperformed OpenAI's ChatGPT in the App Store rankings. This caused the tech company's shares to plummet, which also resulted in a major cyber attack on DeepSeek's servers.
OpenAI's allegations of copyright infringement
OpenAI has also faced multiple allegations of copyright infringement from newspapers and content creators, including lawsuits from The New York Times and well-known authors. They accused the company of training its models on their articles and books without permission.