Lesson 2 of 6
The training-data fight
6 min read
Before an AI could paint in your style, it studied millions of real images — maybe including yours. Was that fair learning, or taking without asking?
Built on human work
AI image and video models don't invent their skill from nothing. They learn from huge piles of human-made work scraped from the web — photos, illustrations, paintings — most of it made by people who were never asked and never paid. When the model makes something new, it's standing on all of that.
AI creativity is built on real people's work — often gathered without consent or payment.
Still being decided
Artists and writers are suing; AI companies argue it's fair use, like a person learning from everything they've seen. Courts in different countries are ruling different ways, and the rules are still moving. Some tools now offer opt-outs, licensed data, or pay-per-use — but there's no settled answer yet.
"Was the training fair?" is genuinely unsettled — treat confident claims on either side with caution.
If you publish or sell AI art, know that the ground is shifting. Prefer tools that are transparent about their training data and offer creators an opt-out. (General explanation, not legal advice.)
The gist
- —AI models learn from vast amounts of human-made work, often taken without consent.
- —Whether that's fair use or infringement is being fought in court, country by country.
- —The rules aren't settled — favour transparent, opt-out-friendly tools.
Why is the way AI image models were trained so controversial?
Continue in the app
Take the whole AI Media: Ownership, Deepfakes & Trust course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.