YouTube warns OpenAI not to use its videos to train Sora: it would be a “clear violation” of its policies

The executive director of YouTubeNeal Mohan, has stated that, in the event that the videos of content creators published on his platform have been used by OpenAI to train its Sora Artificial Intelligence (AI) model, the company would be carrying out a “clear violation” of YouTube policies.

Sora is an AI model that OpenAI presented last February, a tool capable of creating realistic video scenes from text instructions, achieving a detailed result, complex camera movement and even introducing multiple characters with emotions .

To generate the videos, Sora creates content from other videos. In this sense, in a recent interview conducted by The Wall Street Journal, OpenAI’s chief technology officer, Mira Murati, indicated that the data they use to train Sora is “publicly available data and licensed data.”

In this context, asked if said data included videos from YouTube, Instagram or Facebook, Murati indicated that she was not aware of it.

Now, the CEO of YouTube, Neal Mohan, has ruled in this regard that, if OpenAI has used the videos from its platform to train the Sora model, it would be a “clear violation” of its policies.

Specifically, as the CEO of the platform has detailed in an interview with Bloomberg, there is currently no confirmed information on whether OpenAI has actually used the content of YouTube videos to train Sora. However, he has ruled that, if it had been done at any time, the company led by Sam Altman would be violating YouTube’s terms of use.

Following this line, Mohan explained that, in part, this is a misuse of the platform’s content because content creators who decide to publish videos on YouTube expect their work to be protected by the terms of service.

“When a creator uploads their hard work to our platform, they have certain expectations,” Mohan added, while specifying that one of these expectations is that the download of their content, such as transcripts or video fragments, is not allowed. “Those are the rules of the road in terms of content on our platform,” he stated.

On the other hand, regarding the use of YouTube videos to train Google’s AI model, Gemini, the CEO of the platform has indicated that it is based on YouTube’s individual contracts with the creators, which detail whether whether the content is allowed to be used to train the AI ​​or not.