The US Copyright Office’s position has been consistent and is now settled at the highest level that’s likely to weigh in: purely AI-generated output isn’t copyrightable under US law, because copyright protection requires human authorship. The Supreme Court declined to hear a challenge to this position in March 2026, leaving the Copyright Office’s rule in place. This isn’t legal advice, and copyright law varies by country, but the US position is worth understanding clearly since it affects a lot of AI-assisted content being published right now.
What’s Actually Not Protected
Content generated by AI with no meaningful human contribution, a prompt in, an output accepted as-is, doesn’t qualify for copyright protection. That means anyone could technically reuse it, since there’s no copyright to enforce in the first place. Simply writing a prompt and accepting the first result is explicitly called out by the Copyright Office as insufficient to establish authorship.
What Can Actually Be Protected
AI-assisted work is protected for the human-created portions of it. The Copyright Office evaluates this case by case, looking at how much a person selected, edited, arranged, or otherwise made genuine creative choices about the final output rather than just accepting generated content wholesale. Meaningful editing, combining AI output with original human-created elements, and making deliberate creative choices about structure, selection, and arrangement all count toward establishing the human authorship that copyright actually requires.
What This Means in Practice
Content that’s generated and published with minimal human involvement carries real risk: it’s not just weaker protection, it may have no copyright protection at all, meaning no legal basis to stop someone else from reusing it. Content that genuinely involves human editing, selection, and creative decision-making is on much firmer ground, and documenting that human contribution (drafts, edit history, notes on creative choices) is useful if the question of authorship ever actually comes up.
Using AI as a drafting tool that a human then genuinely edits and shapes is different, legally, from publishing raw AI output. That distinction is exactly what the case-by-case analysis turns on.
The Practical Takeaway
If content matters enough to want copyright protection over it, actual human editing and creative decision-making, not just prompting and accepting, is what establishes that protection under current US law. This is a genuinely evolving area, the specific line the Copyright Office draws has shifted before and may shift again, so treating any specific rule as permanent is a mistake, but the core requirement, real human authorship, has held consistently through every ruling so far.