There is a line in most AI platforms’ T&Cs that states the output belongs to the user. But how long will that ownership last? And how much will it be worth once a campaign goes live across multiple markets with GenAI content? Here’s what global marketers need to know.
There is the legal reality of ownership, that most AI platforms’ terms assign whatever rights exist in the output to the user, but whether copyright exists at all is a separate question for that particular piece. But there is also the practical assumption that comes with it, that owning something means you can use it anywhere, for anything. With AI-generated content, that second definition does not hold. The copyright itself may be weaker than it looks, and a growing set of labelling and disclosure rules governs how and where you use the work. So the real question is not only whether you own it, but how much that ownership actually lets you do.
Generating content does not automatically mean you own the copyright: supplier terms, local laws and client contracts determine who does. Even then, it does not stop someone else, using a similar prompt and working in a similar way, arriving at an extremely similar output. Even where you hold rights in your output, they’re unlikely to extend to similar work someone else generates independently. Global regulations underpinning this principle have been consistent so far, for example:
In March 2026 the US Supreme Court declined to revisit the ruling that copyright requires human authorship.
The same month, the UK government proposed removing the protection that has existed since 1988 for wholly computer-generated works, though without a timetable, and nothing has been published since.
South Korea reached a similar position in official guidance last year, confirming that copyright cannot be registered for output produced without human creative intervention.
Three markets, three routes, but the same destination: where there is no meaningful human input, there may be no copyright.
Meanwhile guidance around disclosure requirements is evolving considerably faster.
In Europe, the EU AI Act’s transparency obligations came into force on 2 August 2026. Depending on the creative, EU rules do not require a visible label on every AI-assisted advert; in some cases an invisible, machine-readable marker is enough.
China has required both a visible label and an embedded technical marker since September 2025.
Since June 2026 in South Korea, anywhere an AI-generated figure appears in an ad, the words “virtual person” must be displayed next to it for as long as it is on screen, and presenting such a figure as recommending something from genuine personal experience is prohibited outright.
These disclosure rules don’t change who holds the copyright, but they change what you can do with the asset once it is live. A mandatory label, where one applies, is a condition on use that you did not choose.
In practice, ownership tends to be narrower than the T&Cs imply. You can hold the copyright in a specific output, but not reliably enough to keep others off something similar, and not freely enough to escape the labelling rules that more markets are starting to introduce.
Six things worth settling before GenAI is used at any scale for imagery, videos, voiceovers or music:
Start with the usage, not the tool: Establish which media, which territories and what usage duration you need, then choose accordingly. The usage requirement should come first and every other decision should follow from it.
Check what your subscription tier actually covers: Generated voice and music can carry different usage rights depending on the tier that produced them, and whether broadcast or cinema usage is covered depends on the licence you’ve purchased, not simply the tier.
Be careful what goes in: using clips, voices or images you don’t own, or aren’t permitted to use as references, can create risk later.
Watch the wording of the prompt: Asking for work in the style of a named artist, or a voice that sounds like a specific person, creates obvious exposure. But leaving the name out does not remove the risk: if the output still imitates their work or likeness, the infringement risk remains.
Keep the records: Prompts, references and project files, held in one place, are what let you show how something was made.
Make sure you know who carries the risk: Ask your suppliers to disclose which tools and models they used on your behalf, and where the potential risk lies.
Purely AI-generated output is not protected the way human-made work is, though the human creative contributions around it, such as editing, selection and compositing, can still have copyright protection. Many providers’ terms are honest about it: they assign copyright to the client, then concede that the protection may not exist, and may not hold against competitors using the same tools. That limits how distinctive your output can be, and how firmly you can hold it as your brand’s own. It also shows where the technology earns its place. GenAI supports the work at volume and speed, wherever your content needs it, but it is a poorer fit where brand consistency, a precise message and a distinct look and feel matter most. Every brand, product and campaign carries its own requirements, and those should decide the rights you secure before anything is generated, not after.
Managing usage rights and disclosure requirements as part of production across markets is part of what we do. If you are producing for global campaigns, see how we handle multi-market delivery here.
This article is for general information only and isn’t legal advice. Rules vary by market and are changing quickly, so take specialist legal advice for your specific situation.