Who Owns AI-Generated Content? UK Copyright Rules Explained

The uncomfortable short answer

UK copyright arises automatically when an original work is created; there is no register. The difficulty with AI-generated content is not registration but authorship: copyright protects the expression of human creativity, and the courts have not yet settled how much human input turns a machine output into a protectable work. So the honest answer to "who owns my AI-generated logo?" is: possibly you, possibly nobody, and the difference depends on how much of you is in it.

That uncertainty cuts both ways. You may hold weaker rights than you assume in content you generate, and you may face claims if a generated image reproduces someone else's protected work. Both risks are manageable with a few habits, covered below.

What UK law actually says

The Copyright, Designs and Patents Act 1988 contains a provision most countries lack. Section 9(3) says that for a computer-generated work with no human author, the author is "the person by whom the arrangements necessary for the creation of the work are undertaken". Such works get 50 years of protection rather than the usual life of the author plus 70.

That sounds reassuring, but there is a wrinkle: copyright also requires a work to be original, in the sense of being the author's own intellectual creation, and it is genuinely unclear how a work with no human author satisfies a human-creativity test. Lawyers disagree, and the point has not been squarely decided by a court.

The practical consequence is a sliding scale. A blog post you prompted, restructured and substantially rewrote is on solid ground as your own work. An image generated from a two-word prompt and used untouched is on the weakest ground. The more human creative input you add, the stronger your position becomes.

The training-data fight, and why it matters to you

The highest-profile UK dispute has been Getty Images v Stability AI in the High Court, over the use of Getty's photo library in model training. The 2025 judgment resolved less than many hoped: Getty's central training claims fell away during the trial, largely because the training happened outside the UK, and Getty succeeded only on a narrow trade mark point about generated images containing its watermark. The big question of whether training on copyrighted work infringes UK copyright remains substantively unanswered here.

Meanwhile the government has consulted on a text-and-data-mining exception that would let AI developers train on copyrighted material unless rights holders opt out, a proposal that drew strong opposition from the creative industries. At the time of writing, the final legislative position is still unsettled.

For a small business the exposure is practical rather than theoretical: the legal risk around training sits mostly with model providers, but if an output you publish reproduces a recognisable copyrighted work, a watermark or a real person's likeness, the complaint lands on you, not the tool.

What the tool's terms actually give you

Most major providers assign whatever rights exist in the output to the user; OpenAI's terms, for instance, work this way. But read the tier you are actually on, because some image tools grant full commercial rights only on paid plans, and free-tier terms can be markedly different.

• An assignment of rights is not a guarantee that rights exist; a provider cannot give you a copyright the law does not create
• Outputs are not exclusive: another user with a similar prompt can get a similar result, and the terms will not stop them
• Some enterprise products, including Adobe Firefly and Microsoft's Copilot offerings, have offered indemnification against copyright claims over outputs, which is a genuine differentiator for commercial use
• Terms change; keep a dated copy of the version you relied on when producing important assets

Logos, blogs and product images: the practical rules

Logos and brand assets

A logo is the one asset where weak ownership really hurts, because the whole point is stopping others from using it. Two protections matter: copyright, which is uncertain for pure AI output as explained above, and trade mark registration with the UK Intellectual Property Office, which protects the mark for your goods and services regardless of who or what drew it. The robust route: use AI for concepts, have a human designer develop the final mark, keep the working files, and register the trade mark.

Blog content

AI-drafted posts you substantially edit are the low-risk case; the bigger issue is accuracy and quality, not ownership. For cornerstone content, keep your prompts and edit history, because they evidence human authorship if it is ever questioned.

Product and marketing imagery

• Do run a reverse image search on hero images before publishing them
• Do not prompt "in the style of" named living artists for commercial assets
• Do not generate images of real people, celebrities or other brands' products
• Do check whether your plan actually permits commercial use before the campaign goes live

Keep records and treat outputs as drafts

The pattern across all of this: human input strengthens your rights, and records prove that input existed. Save prompts, iterations and edited versions for any asset your business depends on. If an agency or freelancer produces work for you using AI, make sure the contract assigns all rights to you, requires them to disclose AI use, and includes a warranty that the output does not infringe anyone else's rights.

This article is general information rather than legal advice; for disputes or high-value contracts, speak to an IP solicitor. And if you want brand assets built to be defensible from the start, our team can help.