Recommended Free Tools
OpenAI COO Brad Lightcap did not say whether Sora was trained on YouTube videos when Bloomberg interviewer Shirin Ghaffary asked him directly at the Bloomberg Technology Summit in San Francisco on May 10, 2024. The exchange, reported by Futurism, left the question unresolved: the public material covered here neither confirms that YouTube videos were used nor establishes that they were excluded.
What Brad Lightcap said about Sora and YouTube
At the May 10, 2024, summit, Ghaffary asked Lightcap: “Can you say, and clear up once and for all, whether Sora was trained on YouTube data?” Futurism’s account says Lightcap began, “Yeah, I mean, look, the conversation around data is really important,” then spoke generally about knowing where data comes from and the possibility of a future “content ID system for AI.” It quotes him later saying, “So, yeah, we’re looking at this problem. It’s really hard.”
As an Amazon Associate I earn from qualifying purchases.
According to the same report, Ghaffary followed up: “So no answer on the YouTube. For now.” These are quotations as reported by Futurism, not a verified complete summit transcript. The article’s headline characterizes Lightcap as having “choked,” but his hesitation or failure to give a yes-or-no answer is not evidence of what Sora’s training set contained.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Had another OpenAI executive answered the question?
No. Futurism also recounts an earlier Wall Street Journal interview with then-OpenAI CTO Mira Murati. She said, “We used publicly available data and licensed data.” When asked whether that included YouTube videos, she replied, “I’m actually not sure about that.” Futurism reports that Murati later confirmed Shutterstock videos had been used; that detail does not resolve whether YouTube videos were used.
What OpenAI disclosed about Sora’s training data
OpenAI’s February 2024 technical report on Sora describes the model’s training in broad terms, including learning from internet-scale data and its technical approach. It does not identify YouTube as a source or provide a source-by-source list of training videos. The Associated Press likewise reported at Sora’s announcement that OpenAI had not disclosed the imagery and video sources used to train the model (AP report).
That lack of disclosure has a limited but important meaning: these sources do not establish whether YouTube videos were included. It is not proof that they were used, and it is not proof that they were not.
YouTube’s rules are a separate question
Bloomberg reported that YouTube CEO Neal Mohan said using YouTube videos to train Sora would violate YouTube’s rules (Bloomberg report). That describes YouTube’s stated position on its rules; it does not establish what data OpenAI actually used to train Sora.
What can be concluded
- Lightcap was asked directly whether Sora was trained on YouTube data, but did not answer yes or no in the exchange reported by Futurism.
- Murati’s earlier comments also left YouTube’s inclusion unconfirmed.
- OpenAI’s published technical material and the AP’s launch coverage do not provide a source-by-source disclosure that settles the question.
- YouTube’s stated rules position is not evidence that OpenAI either used or avoided YouTube videos.
Based on these sources, the accurate answer to “Was Sora trained on YouTube videos?” is that it remains unknown.
Quick Recap
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
