Who Owns AI-Generated Content?

Five questions, not one

1. 1What rights does the platform grant the user? Most major AI platforms assign or license output to the user in their terms — but a grant of rights between you and the platform says nothing about the rest of the world.
2. 2Does copyright arise in the output under applicable law? UK law (Copyright, Designs and Patents Act 1988, section 9(3)) contains a provision for computer-generated works with no human author, treating the person who made the arrangements as the author — but its scope is debated, and the UK government has been consulting on copyright and AI. Whether a largely machine-generated output attracts enforceable copyright is not a settled, universal yes.
3. 3Could the output infringe somebody else's rights? Generated text, images and code can reproduce or closely resemble existing protected works, trademarks or distinctive styles. This risk exists whatever your platform terms say.
4. 4Can the business prove sufficient human creative contribution? Works with substantial human authorship — selection, arrangement, editing, creative direction — stand on much firmer ground than raw generations. Keep records of the human work.
5. 5Can a brand asset be registered or protected? Logos and marks generated by AI may face questions about originality and prior similar outputs. For assets the business will depend on, involve a trademark professional before committing.

Prompts, sources and outputs are different things

Your prompts may themselves contain protected material — pasting a competitor's brochure or a client's document into a tool raises its own questions before any output exists. Source material you upload remains subject to whatever rights and confidentiality attach to it. And outputs inherit risk from both: an output shaped by someone else's protected input is not cleansed by passing through a model.

Training-data disputes: watch, don't assume

High-profile litigation over the use of copyrighted works in AI training — including Getty Images' proceedings against Stability AI in the English High Court — has produced early decisions on some issues while broader questions remain contested, and the UK government has consulted on how copyright law should treat AI training. Do not treat ongoing litigation or open consultations as settled law in either direction. What a business can do today is practical: choose reputable platforms, prefer tools with commercial-use terms and provenance commitments, and keep humans meaningfully involved in work that matters.

Practical rules for business content

• Check the current terms of the specific platform and plan you use — rights differ between free and paid tiers and change over time.
• Keep records: prompts, drafts, editing history and who did what. Provenance is your evidence of human contribution.
• Run trademark and similarity checks before adopting AI-generated names, logos or distinctive brand assets.
• Avoid prompting for a living artist's or competitor's style for commercial work — style imitation invites disputes even where the law is unsettled.
• For client work, agree in the contract who owns deliverables and whether AI assistance is acceptable and disclosed.
• Consider disclosure where authenticity matters to the audience — and check sector rules, but note there is currently no blanket UK requirement to label every AI-assisted business output.