YouTube CEO: OpenAI may have violated platform rules to train Sora

Speaking to Bloomberg, Mr. Mohan said this would be a 'clear violation' of YouTube's Terms of Service (ToS). This is the first time the YouTube CEO publicly spoke on this topic.

Mr. Mohan added that he had no information on whether OpenAI used YouTube videos to train Sora. However, if that's the case, this would go against YouTube's policy.

It's worth noting that when OpenAI CTO Mira Murati spoke to The Wall Street Journal's Joanna Stern earlier, she dodged the question of whether Sora was trained based on YouTube videos. Murati said, 'I'm really not sure about that' and declined to elaborate further saying 'I won't go into details about the data that was used'.

The YouTube CEO added that Google's Gemini AI model itself complies with YouTube's individual policies and contracts with creators. AI models can take channel names, video titles or creator names but are not allowed to download or train on 'recordings' or 'video bits'.

Mohan added that Google signs a licensing contract with each creator, and some YouTube video stores can be used to train Google's AI models.

OpenAI has been facing challenges around intellectual property rights for some time to train AI models. New York Times sues OpenAI for copyright infringement. The Intercept also recently filed a lawsuit against OpenAI. Now, we'll have to wait and see how OpenAI responds to accusations of violating YouTube's rules if the company has indeed trained its Sora model on YouTube videos.

Jessica Tanner

Update 05 April 2024