brandleys
← All insights Content and copyright

Generated material in your brand, and what you can rely on

Most businesses now have generated material somewhere in what they sell: web copy, product imagery, marketing assets, sometimes the logo itself. The tools were quick and the results were usable, so nobody paused to ask what the company actually owns. In the United Kingdom that answer is less settled than the confidence with which it is generally assumed.

Copyright in the United Kingdom arises automatically, without any registration, in original works created by an author. Originality is generally understood to require that the work is the author's own intellectual creation, which is a question about human choice rather than about effort, skill or quality of the result. Where the human contribution consisted of a short instruction and the decision to accept what came back, whether that person has created anything protectable is genuinely open.

This uncertainty is comfortable to live with, because for long stretches nothing tests it. Generated copy sits on a website for years and nobody asks who authored it. The question arrives from outside, and it tends to arrive at speed: when a competitor lifts the asset, when a platform requires the business to certify that it owns what it is complaining about, or when somebody buying the company asks for the chain of title to the brand.

What United Kingdom law protects, and who it treats as the author

The starting point is that copyright protects the expression a human author created, and belongs to that author. Editing, arranging, selecting and reworking generated material are all human acts, and material that has been through genuine creative work by a person stands in a considerably better position than material that was accepted as it came. What matters is whether a person made real creative choices about the thing that was published, and whether anyone in the business could now describe what those choices were.

United Kingdom legislation also contains a long-standing provision addressing works generated by a computer where there is no human author, treating as the author the person by whom the arrangements necessary for the creation of the work were undertaken. It was drafted well before the current generation of tools existed. How it applies to a workflow of prompting and accepting has not been resolved by the courts, the protection it confers is narrower than that given to a work with a human author, and whether the provision should remain in its present form has been the subject of government consultation. A business relying on it is relying on something untested, which is a different thing from relying on nothing, and a very different thing from relying on a settled right.

What the tool's terms actually give you

Between the business and the provider, the contract governs. The terms vary a great deal and they change. Some purport to assign to the customer whatever rights exist in the output. Some grant a licence and keep ownership. Some reserve the right to use inputs and outputs to develop the service. Terms often differ between the free tier and the paid one, and between the consumer application and the business product, which matters because teams frequently start on one and never formally move.

The limit on all of this is simple enough. A provider can only give what it has. Where no copyright subsists in an output, a clause assigning ownership of that output transfers an empty right. The clause is still worth having, because it tells you the provider is not asserting a claim over your asset. It does not create exclusivity against anyone else in the world.

Several providers now offer to stand behind their output if a third party complains. Those offers are typically conditional on the tier used, on particular settings being enabled, and on the customer not having steered the tool towards a specific existing work, and they sit under the liability caps in the same agreement. An indemnity addresses what happens when somebody complains about you. It does nothing about whether you can complain about somebody else.

Assignments that may transfer very little

Businesses commission freelancers and agencies for creative work and take an assignment of the rights. That practice is correct and it should continue. It has quietly become less protective than it was.

An assignment transfers whatever the assignor holds. Where the deliverable was generated, what they hold may be a good deal less than either party assumed when the document was signed, and in some cases may be very little at all. Nobody discovers this at signature, because signature is not a test of anything. It is discovered at the point the business needs to enforce, which is the worst possible moment to learn that the paperwork was in perfect order and the underlying right was thin.

The practical consequence is that what a business buys from a supplier of creative work now includes a representation about how the work was made. Most contracts still do not ask, and most suppliers are not volunteering.

Where the question gets tested

Enforcement is the first place. Platform brand protection and takedown routes require a complainant to identify the right being relied on and to certify entitlement to it, often in terms that carry consequences for inaccuracy. A business asserting copyright in an asset whose authorship it cannot describe is in an uncomfortable position before it has even started, and the party copying the asset may be in a stronger one than it appears.

Diligence is the second. A buyer asking for the chain of title to brand assets is asking about the exclusivity being paid for. Where the trail runs into generated material with no describable author, the buyer's concern is whether the asset can exclude a competitor at all, since exclusivity is a large part of what a brand premium is paid for.

Both are made harder by a records problem that is entirely ordinary. Content teams rarely log which tool produced which asset, or the point at which a person took over and what they changed. By the time the question is asked, the material sits in a shared drive with no history, the workflow has since been replaced, and the person who could have described it has left. Reconstructing that account after the fact is the expensive part, and it is expensive in exactly the cases where the asset turned out to be valuable.

The exposure running the other way

Ownership is only half of the position. Whether using copyright works to train a model requires permission, and on what terms, is contested in the United Kingdom. It has been litigated in England and Wales and it is the subject of active government policy work, so anything written about it should be read as a snapshot. The United Kingdom exception permitting text and data mining is presently narrow and directed at non-commercial research, and proposals to widen it have been consulted on without settled resolution.

For a business that merely uses the tools, the practical exposure is more immediate than the training debate. An output can reproduce a recognisable part of an existing work, or come close to a registered trade mark or a registered design, and the business publishing it carries that risk whatever the provider's terms say about ownership. Style in the abstract is not generally protected by copyright in the United Kingdom, but a particular expression is, and the line between them is a question of fact rather than a rule that can be applied by a marketing team on a deadline.

Which produces an uncomfortable pattern. The exposure concentrates in the most public and most valuable assets, and those are frequently the ones produced fastest, because the deadline was tight and the tool was there.

What is not worth doing

Banning the tools achieves very little inside an operating business. Teams use them anyway and stop mentioning it, which removes the only genuinely useful thing management had, namely knowing where generated material sits.

Auditing every generated asset is disproportionate and nobody finishes it. Most content is ephemeral, is never copied, and would never be worth enforcing if it were. Where a business would not spend anything to stop somebody reproducing an asset, the ownership question about that asset is academic and can be left alone permanently.

The sorting that repays effort is narrow: the assets the business would actually defend, and everything else. Names, logos, product identity, signature imagery, anything filed at the Intellectual Property Office or printed on packaging. For most businesses that list is short, and drawing it up is an internal conversation that requires no professional involvement at all.

There is also rarely an emergency. A business that has already built its brand on generated material is not in trouble today by reason of having done so. The moment to look is before a filing, a launch, a raise or a sale, when the answer can still change what gets done.

The mistake to avoid

Adopting a generated logo or product identity across packaging, filings and every customer-facing surface, and asking the ownership question only when a copy appears. At that point the asset carries the goodwill of the business, the copyist is trading on it, and the two available responses are an enforcement route that may rest on a right nobody can describe, or a rebrand of something customers already recognise. Both cost more than commissioning the asset properly would have done.

This guide is general information about how these matters usually run. It is not advice, and nothing becomes advice until terms are agreed in writing. Brandleys Legal Ltd delivers reserved legal activities alongside regulated partners.

Free check, about a minute

IP gap audit

What you own but have not protected, in plain English.

Run it ›

Before it becomes the brand

If generated material is heading for a logo, a filing, a launch or a diligence pack, the useful conversation happens while the asset can still be changed cheaply. Tell us where it sits and what it has to do.