OpenAI’s new video tool, Sora, is turning heads with how realistically it can generate short clips that look like they came straight from Hollywood. But there’s a catch. Some tests suggest that Sora might have been trained on copyrighted material—things like Netflix shows, TikTok videos with watermarks, and even logos from video games. This raises a big question: Did the AI learn from content it wasn’t allowed to use?

When people ask Sora to create a trailer or a clip, it sometimes produces something that looks suspiciously like popular movies or shows. For example, it might generate a scene that resembles the opening of a well-known Netflix series or a recognizable animation style. That’s not just coincidence. It’s probably because the AI learned from watching similar content during its training. Researchers say that scraping videos from platforms like YouTube has been common for AI developers, even if it’s technically against the rules.

Are Platforms Like YouTube Allowing Copyright Violations?

Platforms like YouTube and TikTok clearly state that you shouldn’t scrape or download videos without permission. Yet, many companies have developed tools that can collect millions of clips quickly. These tools are used to train AI models, helping them learn how to generate images, videos, or music. Companies like Nvidia and Runway ML are reported to use such methods, turning these platforms into huge data sources—sometimes without proper consent.

This practice sparks a debate about legality and ethics. Some experts compare it to borrowing a book from the library—using content for learning is fair in some cases. Others argue it’s unfair because it takes creative work without paying or asking for permission. A group of YouTubers has already sued OpenAI, claiming the company used millions of hours of their videos and transcriptions without approval. The outcome of these lawsuits could shape future AI development and copyright laws.

What Could This Mean for Creators and the Future of AI?

If AI models like Sora are trained on copyrighted works without permission, it could hurt the people who create original content. Imagine a small filmmaker or animator who spends months or years making their work. If an AI can produce similar content instantly, what happens to their income and careers? It’s a tricky balance. On one hand, AI could help small creators by giving them new tools to tell stories or make art. On the other, it threatens to replace jobs and devalue the effort behind original content.

OpenAI claims it only trains Sora on “publicly available and licensed data,” but the results suggest otherwise. Ethicists like Margaret Mitchell from Hugging Face point out that the bigger issue isn’t just about legal definitions. It’s about respecting the choices of creators and whether they should have control over how their work is used. The legal battles underway could change what “fair use” means in the world of AI.

If courts decide that scraping copyrighted videos is okay, we might see a flood of AI-generated movies, games, and ads that feel familiar but don’t belong to anyone. That could lead to a future where original content is harder to protect, and everyone—artists, companies, and consumers—must rethink what value creativity truly has. The stakes are high, and the outcome will influence how AI develops in the coming years.

Inspired by

• https://ai2people.com/openais-sora-sparks-copyright-storm-did-ai-learn-from-netflix-and-tiktok-without-permission/