LANDMARK · 7 MIN · JUDGE
Copyright and permission
After this landmark, you can separate access, transformation, attribution, and permission when deciding whether an AI-assisted use is safe.
Generative AI blurs a distinction that copyright keeps sharp: being able to reach a work is not permission to reuse it. It helps to separate four different things. Access is whether you can see or obtain the material. Transformation is how much you change it — a summary, a paraphrase, a derivative, or a near-copy. Attribution is whether you credit the source, which is about honesty and academic norms, not permission. Permission is whether the rights-holder has actually licensed the use, which is the one that determines legality. A model can output text or images that closely track copyrighted works it was trained on, and it won’t tell you when it does. On top of that, the law itself is unsettled — whether training on copyrighted data is infringement, and who owns purely AI-generated output, are being actively litigated and legislated, and the answer differs by country. So the honest position is caution, disclosure, and not treating fluent output as automatically clear to use.
This is the canonical concept. It stays the same across learner lenses so personalization never changes the underlying facts.
◌ EVERYDAY LENS
What this looks like for you
Don’t assume AI output is yours to publish freely, and don’t assume it’s original — it can closely echo existing work. For anything you’ll share publicly, check it isn’t reproducing a recognisable song, image, or passage, and credit real sources you drew on.
MAKE A DECISION
A designer uses AI to generate a logo that looks strikingly like a famous brand’s mark, planning to use it commercially. What’s the sound judgment?
CARRY THISTake one AI output you plan to reuse and sort it across access, transformation, attribution, and permission.