OpenAI claims that Chinese start-up DeepSeek used its technology to build a rival AI model. The company says it has proof that DeepSeek may have violated its policies by using a technique called "distillation." This method helps smaller AI models improve by learning from larger ones.
![DeepSeek vs ChatGPT DeepSeek vs ChatGPT](https://data.ibtimes.sg/en/full/79491/deepseek-vs-chatgpt.jpg?w=736)
A source close to OpenAI told the Financial Times that DeepSeek copied its technology to create its own product. OpenAI has strict rules against using its tools to develop competing AI models. However, it has not publicly shared evidence to support its claims.
The U.S. has long worried about intellectual property theft from China, especially in AI and biotechnology. Some experts believe DeepSeek's rapid rise raises questions about how it achieved such success with limited resources. Its new AI model, R1, has performed as well as leading U.S. models despite a much smaller budget.
The controversy has sparked concerns beyond OpenAI. The U.S. Navy has reportedly banned DeepSeek over fears that China could use it to access sensitive data. Tech investors are also reacting. On Monday, Nvidia, a major AI chip supplier, saw its stock plunge 17%, wiping out $589 billion in value. The market later rebounded, with Nvidia's stock recovering 9% on Tuesday.
DeepSeek's success has surprised many in the industry. The company claims it trained its AI model using only 2,048 Nvidia H800 graphics cards and spent $5.6 million. This is far less than what firms like OpenAI and Google invest in similar projects. Some experts believe DeepSeek may have trained its model using OpenAI's GPT-4, which would violate OpenAI's terms.
Last year, OpenAI and Microsoft investigated accounts linked to DeepSeek. These accounts allegedly misused OpenAI's technology, leading to their termination. Bloomberg News first reported the investigation.
AI and tech leaders are speaking out. David Sacks, an AI advisor to former President Trump, suggested DeepSeek may have copied OpenAI's work. "There is strong evidence that DeepSeek used OpenAI's models," he told Fox News. However, he did not provide proof.
OpenAI says it is working with the U.S. government to stop companies from stealing its AI technology. It also continues to defend its intellectual property. Meanwhile, irony surrounds the situation, as OpenAI itself faces lawsuits for allegedly using copyrighted material without permission. The New York Times and other publishers have sued OpenAI, accusing it of training ChatGPT on their work.
Despite these challenges, OpenAI has struck deals with major publishers, including News Corp, Condé Nast, and Le Monde. These agreements allow OpenAI to access news archives legally, potentially shaping the future of AI and journalism.