Financial Times: OpenAI says it has evidence that DeepSeek has engaged in infringement.

The violent shockwaves triggered in the U.S. artificial intelligence industry by Chinese tech company DeepSeek have persisted for several days, and some dissenting voices are continuously emerging. The London Financial Times reported on Wednesday (January 29) that OpenAI, the developer of the U.S. AI chatbot ChatGPT, stated it has found evidence that Chinese startup DeepSeek used OpenAI's proprietary models to train its own open-source chatbot. This provides new corroboration for those who suspect the Chinese company may have infringed on intellectual property rights. The Financial Times said that the ChatGPT maker, headquartered in San Francisco, stated it has seen some evidence of infringement by DeepSeek developers through "distillation" technology. The so-called "distillation" technology refers to a practice where developers use the output of more powerful models to train smaller models, thereby achieving higher performance at lower cost. The Financial Times said this is a common practice in the industry, but the problem is that DeepSeek's intelligent chatbot is a competing product in the same category as OpenAI's ChatGPT, and this practice allegedly violates OpenAI's product usage terms and may constitute infringement. The report quoted a person close to OpenAI as saying, "The problem is you [take it out of the platform] to create your own model for your own purposes." Tesla CEO Elon Musk has also reportedly pointed out recently that he does not believe DeepSeek's claimed AI R&D capabilities and suspects that the Chinese company possesses far more Nvidia graphics cards than it has disclosed. The tech news site TheVerge.com published an article on Wednesday saying that OpenAI and U.S. tech giant Microsoft are investigating whether this Chinese competitor used OpenAI's API (Application Programming Interface) to integrate OpenAI's AI models into DeepSeek's own models. Sources said that Microsoft security researchers discovered at the end of 2024 that a large amount of data was leaked through an OpenAI developer account, which Microsoft believes is linked to DeepSeek. The Financial Times said OpenAI declined to comment further or provide detailed evidence. The company's product usage terms clearly stipulate that users cannot "copy" any of its services or "use its outputs to develop models that compete with OpenAI." DeepSeek's released R1 reasoning model surprised the market, investors, and tech companies in Silicon Valley. This model is low-cost, highly ranked, and comparable in capability to leading U.S. models. After DeepSeek's model was released last weekend, it triggered a sharp drop in U.S. tech stocks on Monday. U.S. AI chip giant Nvidia's stock fell 17%, with its market value evaporating nearly $600 billion. Investors worry that the U.S. model of developing AI through massive investment may have significant bubbles, which could lead to a reassessment of their value. However, on Tuesday, market anxiety eased, with the Nasdaq rebounding over 400 points and Nvidia's stock also rebounding nine percentage points. Source: https://www.voachinese.com/a/openai-says-it-has-evidence-china-s-deepseek-used-its-model-to-train-competitor-20250129/7955346.html