The UK's Financial Times reported late today (29th) that OpenAI has revealed evidence that DeepSeek is suspected of using OpenAI's proprietary models to train its products, possibly infringing on OpenAI's intellectual property. (Background: DeepSeek launches AI multimodal open source model 'Janus-Pro', image generation surpasses DALL-E 3, Stable Diffusion) DeepSeek, a Chinese startup, recently launched an open source inference model R1, which quickly ignited the entire AI technology field. With its 'low cost, high performance' characteristics, DeepSeek has impacted large US tech companies that dominate the AI industry. The stock prices of leading US AI giants such as NVIDIA and OpenAI have plummeted. Even US President Trump has called DeepSeek the 'alarm bell' of the US AI industry. DeepSeek suspected of infringing on OpenAI's intellectual property? However, according to the UK's Financial Times, OpenAI has revealed evidence that DeepSeek is suspected of using OpenAI's proprietary models to train its products, possibly infringing on OpenAI's intellectual property. OpenAI also added that DeepSeek may have used a technique called distillation, which allows DeepSeek to improve its performance using OpenAI's models without incurring high costs for training its models. Distillation techniques are common in the industry, and OpenAI also provides legitimate ways for developers to perform such operations. However, if you use this technique to create competing products for OpenAI, the nature of the matter is different. In addition, AI Tsar David Sacks, appointed by US President Trump, also commented on this incident, believing that DeepSeek may be suspected of infringing on intellectual property. Microsoft assists OpenAI in investigating DeepSeek. In addition, insiders revealed that Microsoft, as a large US tech stock company, has been investigating accounts suspected to belong to DeepSeek in collaboration with OpenAI since last autumn. They have already discovered that DeepSeek used OpenAI's API for distillation. As a result, OpenAI banned DeepSeek's account for violating its terms of use. Currently, OpenAI has issued a statement saying that it will take countermeasures to protect its intellectual property: We know that Chinese companies and others are constantly trying to steal the models of top US AI companies using distillation techniques. We will take countermeasures to protect intellectual property rights, including deciding carefully which cutting-edge capabilities should be included in the released models. We believe that cooperation with the US government in the future is crucial. We need to protect the most powerful AI models and prevent them from being stolen by competitors. Difficulties in protecting OpenAI's rights? However, it should be noted that since distillation is very common in the training of AI models, banning such behavior may also encounter various challenges. For example, Ritwik Gupta, an AI PhD from the University of California, stated: Startups and academic institutions often use commercialized large models that have been aligned by humans, such as ChatGPT, to train their own models, which means that you can obtain research results of top models at a low cost. I am not surprised that DeepSeek may adopt this approach. If they do, it will be very difficult to completely prevent such behavior. In addition, the Financial Times also added that OpenAI itself is facing copyright infringement accusations, including lawsuits filed by The New York Times and several well-known authors, accusing OpenAI of using press releases and book content without permission to train its AI models. Related reports: Zuckerberg warns: Chinese AI models like Deepseek are too powerful, US companies and government should fully block it. How did the H100 rental price plummet 70% and the AI computing bubble burst? DeepSeek shatters the protective moat of the US AI industry, but is it a big plus? Is there something fishy behind the GPU computing power? The article 'OpenAI has evidence of 'DeepSeek infringement', training Chinese AI using GPT distillation technology' was first published on BlockTempo, the most influential blockchain news media in the blockchain industry.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
OpenAI có bằng chứng "vi phạm DeepSeek ", chiếm dụng công nghệ chưng cất GPT để đào tạo AI Trung Quốc
The UK's Financial Times reported late today (29th) that OpenAI has revealed evidence that DeepSeek is suspected of using OpenAI's proprietary models to train its products, possibly infringing on OpenAI's intellectual property. (Background: DeepSeek launches AI multimodal open source model 'Janus-Pro', image generation surpasses DALL-E 3, Stable Diffusion) DeepSeek, a Chinese startup, recently launched an open source inference model R1, which quickly ignited the entire AI technology field. With its 'low cost, high performance' characteristics, DeepSeek has impacted large US tech companies that dominate the AI industry. The stock prices of leading US AI giants such as NVIDIA and OpenAI have plummeted. Even US President Trump has called DeepSeek the 'alarm bell' of the US AI industry. DeepSeek suspected of infringing on OpenAI's intellectual property? However, according to the UK's Financial Times, OpenAI has revealed evidence that DeepSeek is suspected of using OpenAI's proprietary models to train its products, possibly infringing on OpenAI's intellectual property. OpenAI also added that DeepSeek may have used a technique called distillation, which allows DeepSeek to improve its performance using OpenAI's models without incurring high costs for training its models. Distillation techniques are common in the industry, and OpenAI also provides legitimate ways for developers to perform such operations. However, if you use this technique to create competing products for OpenAI, the nature of the matter is different. In addition, AI Tsar David Sacks, appointed by US President Trump, also commented on this incident, believing that DeepSeek may be suspected of infringing on intellectual property. Microsoft assists OpenAI in investigating DeepSeek. In addition, insiders revealed that Microsoft, as a large US tech stock company, has been investigating accounts suspected to belong to DeepSeek in collaboration with OpenAI since last autumn. They have already discovered that DeepSeek used OpenAI's API for distillation. As a result, OpenAI banned DeepSeek's account for violating its terms of use. Currently, OpenAI has issued a statement saying that it will take countermeasures to protect its intellectual property: We know that Chinese companies and others are constantly trying to steal the models of top US AI companies using distillation techniques. We will take countermeasures to protect intellectual property rights, including deciding carefully which cutting-edge capabilities should be included in the released models. We believe that cooperation with the US government in the future is crucial. We need to protect the most powerful AI models and prevent them from being stolen by competitors. Difficulties in protecting OpenAI's rights? However, it should be noted that since distillation is very common in the training of AI models, banning such behavior may also encounter various challenges. For example, Ritwik Gupta, an AI PhD from the University of California, stated: Startups and academic institutions often use commercialized large models that have been aligned by humans, such as ChatGPT, to train their own models, which means that you can obtain research results of top models at a low cost. I am not surprised that DeepSeek may adopt this approach. If they do, it will be very difficult to completely prevent such behavior. In addition, the Financial Times also added that OpenAI itself is facing copyright infringement accusations, including lawsuits filed by The New York Times and several well-known authors, accusing OpenAI of using press releases and book content without permission to train its AI models. Related reports: Zuckerberg warns: Chinese AI models like Deepseek are too powerful, US companies and government should fully block it. How did the H100 rental price plummet 70% and the AI computing bubble burst? DeepSeek shatters the protective moat of the US AI industry, but is it a big plus? Is there something fishy behind the GPU computing power? The article 'OpenAI has evidence of 'DeepSeek infringement', training Chinese AI using GPT distillation technology' was first published on BlockTempo, the most influential blockchain news media in the blockchain industry.