Microsoft examines whether Deepseek incorrectly uses Openai API

Just a few hours after David Sachs firm Deepseek uses Openai models to train their own models, Bloomberg Law reports This Microsoft examines the use of Deepseek from the OPENAI (API) application programming interface.

According to security researchers working for Microsoft, the Chinese company behind A model for reasoning r1 He may have made a large amount of data using Openai’s API in the fall of 2024. Microsoft, which is also the largest OPENAI shareholder, notify Openai of the suspicious activity.

While anyone can register and access the OPENAI API, the company service conditions provide that you cannot use the output to train a new AI model.

“Forbid you from (…) to use production to develop models that compete with Openai,” the company wrote in its Terms of UseS In addition, the company says you cannot “automatically or program (retrieve) data or output”.

The main problem seems to come from distillation, a method used by AI developers to extract knowledge from another model using the behavior of a teacher. It will be interesting to see if Deepseek has found innovative ways to circumvent the speed limit of Openai and to ask for its API on a scale. If this is the case, there will certainly be legitimate consequences.