Just a few hours after David Sacks alleged Deepseek used the models from Openai to train its own models, Bloomberg Law Report That Microsoft Deepseek’s use of the Openai application programming interface (API) examines.
According to security researchers who work for Microsoft, the Chinese company is behind the R1 argumentation model A large amount of data may have sanded in autumn 2024. Microsoft, who is also Opena’s largest shareholder, was informed about the suspicious activities.
While everyone can register and can access Openais API, you must imagine the company’s terms of use that you cannot use the output to train a new AI model.
“It is forbidden to develop the output to develop models that compete with Openaai,” writes the company in its Conditions of use. In addition, the company indicates that you cannot (extract) data or output automatically or program -controlled.
The core problem seems to come from the distillation, a method used by AI model developers to extract knowledge from a different model using a kind of teacher student. It will be interesting to see whether Deepseek has found innovative ways to avoid Openai’s tariff boundaries and query his API on a scale. If this is the case, there will certainly be legal effects.
