Openai points to Deepseek for using distillation to train its AI models based on the chatgpt

Openai has pointed out that he has found evidence that DeepSeek has used Your artificial intelligence models (AI) to train your open source model through the technique of distillation which can cause intellectual property problems.

Deepseek’s AI currently leads download lists in application stores App Store y Google Play Store with your free generative assistant. Also, its new ones R1 Reasoning Models They have generated a great impact on the industry, with responses from companies such as NVIDIA and OpenAi, valueing the progress in the offered.

However, the company led by Sam Altman has now pointed out that, after carrying out an investigation, they have found Evidence that Deepseek has used OpenAi models to train their own models open source.

This has been detailed by sources related to the company in statements to Financial Times, who have indicated that evidence has been identified that suggests that Depseek has used a technique called “distillation” to train their models.

Specifically, distillation is a practice of ‘machine learning’ that allows transmit learning from one model from AI to another. Thus, it allows a smaller model (student model) to imitate the behavior of larger neural networks (teacher model), learning their predictions and using their data.

In this sense, although it is a usual practice in the sector, OpenAI has stated that, at the moment in which these data are used outside the platform, it can be a violation of the terms of the company, where it is detailed that It is not allowed to use the results of their models to “develop models that compete with those of Openai”.

“The problem is when you take it out of the platform and do it to create your own model for your own ends,” explained one of the sources in the aforementioned environment.

In addition to all this, it must be taken into account that, as Bloomberg has collected, OpenAI and Microsoft They were already investigating during the past fall to some accounts that were using OpenAi API to carry out distillation With your models and filter your data, violating the terms of the service. Specifically, although these accounts were blocked, they are believed to be Deepseek owned.

On the other hand, in a statement also collected by Financial Times, Openai has shared that they are aware that China -based companies “They are constantly trying to distill the models of the main companies of the United States“.

In this regard, Technological has underlined the importance of working in collaboration with the United States government for “better protect the most capable models of adversaries’ efforts and competitors to seize American technology. ”