The analogy

Imagine taking a photo with an automatic camera pointed at random out of a train window: the device does everything, you've only pressed. Now imagine a photographer who chooses the light, waits for the instant, composes the frame: the same camera, but the result bears their signature. Copyright (the law that assigns ownership of what one creates) has always looked at the hand and the head of whoever creates, not at the tool. AI is a very sophisticated machine, but it remains a machine: if you merely press, the result is no one's; if you compose, choose, rework, the signature becomes yours again.

How it really works

The rule that holds today in the United States and in Europe is human authorship: to be protected by copyright, content must be born of a creative act by a person. In the United States, in March 2026 the Supreme Court let stand the decisions denying copyright to works without a human author; the Copyright Office continues to register only the parts created by people. The European Union applies the same principle.

From this descends a scale. An image or a text generated only from a command, taken exactly as they come out, have no protected owner: anyone can republish them. Content in which you made substantial creative choices (you reworked, combined, modified significantly) is protected for the part you added yourself. In between there's a gray area assessed case by case, on how much the human hand affected it.

Watch out for something different from ownership: liability and the rights of others. Even if a generated image is "yours," it might resemble an existing work or reproduce a trademark, and there it's you who answers for the possible infringement. Owning the result and having the right to use it everywhere are not the same thing.

What you can do in practice

If you want the content to be yours, leave your mark. Don't stop at the first result of the command: rework, choose, combine, modify. The more your creative contribution is visible, the more solid your ownership of the part you added.

Read the terms of use of the tool you use. Companies can't grant you a copyright the law doesn't recognize, but they set out in their terms what they allow you to do with what you generate (commercial use included or not). Legal ownership is one thing, the provider's permission to use is another: both count.

For uses that weigh (a logo, a cover, content you earn from), keep track of your contribution. Saving the versions and the changes you made is the proof of your creative contribution, useful if one day you have to demonstrate what's yours.

A common misunderstanding

The most widespread misunderstanding is thinking: "I generated it with my account, so it's mine like when I write a sentence." The account and the command aren't enough. The law doesn't look at who paid the subscription or pressed enter, it looks at who performed the creative act. Writing a command, even an elaborate one, is often considered an instruction, not a creation: it's like commissioning a painting by describing it in words, without painting it. That's why a generated image used raw can be copied by others without you having any footholds: not because it was stolen from you, but because it had never become yours.

Frequently asked questions

If I wrote the command myself, is the image mine?

The command alone normally isn't enough to make you the owner. Current law considers the command an instruction given to the machine, not a creative act like painting or writing. Ownership arises from the creative work you add to the result, not from having requested it.

Does the company providing the AI own what I generate?

No, serious companies don't claim ownership of the outputs and in their terms they let you use them, often even commercially. But "I let you use it" doesn't mean "you're the protected owner": the provider's permission and copyright are two different planes. Always check the terms of the specific service.

If AI content belongs to no one, can I use what others generated?

On the copyright plane, a purely generated and unreworked output often isn't protected, so you wouldn't be infringing ownership. But it's slippery ground: it could contain protected parts, reproduce trademarks or existing works, or be covered by the terms of use of whoever generated it. The prudent remedy is not to take for granted that "free of copyright" means "usable without risks": for important uses, verify and, if needed, ask for a legal opinion.