OpenAI’s 2025 Media Manager: A Real Opt-Out or Just Optics?

OpenAI knows the copyright questions aren’t going away, so they’re trying to build an answer. The company just outlined a new tool called Media Manager, slated for a 2025 launch, that’s designed to let creators and content owners specify exactly which of their copyrighted works they want excluded from AI training. It’s a direct response to the growing chorus of artists, writers, and publishers who argue that scraping the web for training data amounts to theft, not fair use. OpenAI frames this as a natural evolution of last summer’s decision to let web publishers block their crawlers via robots.txt, a move they pioneered for the AI industry. But they’re also candid about the limits of that approach, acknowledging that many creators don’t control the websites where their work ends up, and that content is constantly remixed and reposted across domains. The solution, they say, requires cutting-edge machine learning to build what they call a first-of-its-kind tool capable of identifying copyrighted text, images, audio, and video across multiple sources and reflecting creator preferences.

What’s notably missing is any real detail on how this identification will work technically, or whether the default setting will be opt-in or opt-out. Without those specifics, the announcement reads more like a statement of intent meant to ease regulatory pressure than a concrete product roadmap. The 2025 target gives them breathing room, but it also leaves a long window where the current, largely opt-out, system remains in place.

OpenAI is careful to frame the entire AI enterprise as a boon for creators, not a threat. The post highlights partnerships with heavyweights like the Financial Times, Le Monde, and Axel Springer to display publisher content in ChatGPT, along with non-profits like Khan Academy. They make a point of stating they are “not professional writers, artists, or journalists” but rather tool builders. The rhetorical stance is clear: they want to be seen as a platform that empowers, not replaces, creative professionals.

Still, the core tension remains unresolved. OpenAI reiterates its belief that training on publicly available data is fair use, a legal argument that’s currently being tested in courts around the world. Building a preference manager is an admission that the legal landscape is uncertain and that a purely technical opt-out via robots.txt isn’t sufficient. Whether Media Manager becomes a genuine industry standard or a complex, burdensome system that few creators actually use is an open question. For now, it’s a promise tied to a two-year deadline, and a lot can change in the courts before then.