Source : https://www.britishcopyright.org/
Source : https://www.britishcopyright.org/
OBSERVATIONS
Florian Koempel on AI
As if the world needed another AI blog. There are now more AI experts than Royal commentators critiquing the lifestyle of the Sussexes.
AI now inserts itself into every creative stage with the enthusiasm of a new intern who hasn’t yet realised most of the job is scanning files. It drafts concepts, imitates styles, and generates images with the confidence of a camera that has never once forgotten to load the film. Post-production—once the realm of specialists who spoke in mystical tones about “warmth” and “air”—now runs on enhancement tools that never sleep, never complain, and never insist that the background “needs more soul.”
Rights holders have unleashed their own systems to track usage, detect infringements, and chase royalties across continents. Entire catalogues and image libraries are now managed by recognition systems that promise efficiency and omniscience, though one should never believe marketing or start-ups pitching their latest Franken-model to investors. And is it intelligent? Let’s not flatter it.
As new models proliferate, the barrier to entry for “being creative” has collapsed. Excellent news for participation; slightly awkward for anyone who believed art or photography required emotion, craft, or at least some form of intoxication. Authenticity becomes irrelevant. No — let’s try being human.
Licensing negotiations between creators and AI developers are increasing — litigation has a wonderful way of encouraging cooperation. Everyone keeps circling the same principles: follow the law, be transparent about datasets, and label AI-generated outputs so viewers know whether the portrait they’re admiring was made by a person or a GPU.
Under the Data (Usage and Access) Act 2025, the Government must explain the economic impact of its chosen AI policy by 18 March 2026. We await this with the usual mixture of curiosity, scepticism, and the faint hope that clarity might make a cameo on the green screen later. It won’t.
AI-generated images are now so polished that many viewers can’t tell whether they’re seeing a human photographer, a machine, or a machine pretending to be a human pretending to be a machine. Consumers claim to prefer human-made work, but they also claim to prefer vegetables to crisps. People lie, as Greg House says.
The legal framework resembles a half-assembled IKEA wardrobe: all the pieces are present, but no one agrees which way up anything goes. We’ve lived through similar chaos before— peer-to-peer for music, early digital manipulation for photography. AI may follow the same arc, unless policymakers decide to “help” the tech sector by weakening copyright, in which case we can all enjoy a nostalgic return to the early 2000s, minus the charm of dial-up noises.
AI models are only as good as the human creativity they copy. Under UK law, that copying is reproduction. The only question is whether anyone can squeeze themselves into an exception. TDM exceptions were designed for researchers, not industrial-scale ingestion. Temporary copying exceptions apply to fleeting, incidental copies; AI training is neither fleeting nor incidental, no matter how creatively one stretches the word “temporary.”
You use human-created photographs, you get permission. Simples.
The copyright question for AI-generated work is actually simple. Copyright protects human creativity, not machine output. If the AI makes something entirely on its own, there’s no copyright. If a human contributes, only that human part is protected. Simples. No PhD thesis or policy briefing required. Please.
AI-assisted work is still fully protected — just as it was when “AI” meant spell check, noise reduction, or a camera that pretended it knew what “auto” meant. Only the human who creates using AI is protected. The machine gets nothing, certainly no rights rights.
Entrepreneurial rights (like films or broadcasts) are different: they don’t require creativity at all. So an AI-generated image or recording could still get those rights, even if no human created the underlying work (the underlying work not being protected). Its commercial value may be limited, of course, because anyone can recreate it — the photographic equivalent of trying to sell a California sunset.
A functioning licensing market needs two things: people willing to license their work and people willing to pay for it. The creative industries have been doing this since forever; they could do it in their sleep. Ethical AI developers already license data, but the voluntary “trust us” approach has delivered neither transparency nor money. Quelle surprise — tech companies prefer to save their cash for sponsoring ballrooms or commissioning flattering biographies of First Ladies. Regulation will have to carry the weight, as usual.
Weakening copyright would simply reward those who ignore it and tilt the field toward the least well-behaved players. Any workable system has to be opt-in, with creators choosing to license their work for training and getting paid according to how much their work actually influences the output — a radical idea, apparently.
Florian Koempel, 11 February 26 (edited by Co pilot)
Florian is a copyright lawyer, lecturer, and non-executive PICSEL board member. The information provided is for general informational purposes only and does not constitute legal advice.
Trevor Jennis McDermott from GAP Photos:
“Licensing Rights Considerations. 101 of what to look out for”
In Rights-managed licensing, the cost of the image license is typically tied to the media, size, duration, print-run/circulation (for physical print) & territory. Each of these factors is an essential a metric that can be used to measure the level of exposure or exploitation of an image.
As a Copyright holder, you can utilise the information about usage to calculate an appropriate license fee.
Over the last few years, specifically in editorial licensing we have found an increasing number of publishers issuing their own “standard terms” / “rights agreements” – these terms are usually written from the perspective of the publisher and their legal team.
Such agreements will often provide the publisher with much wider rights than the project/product actually requires and use wording that is quite ambiguous as to the final use. Often catch-all phrases or wording, that on the surface sounds quite reasonable, can actually result in open-ended broad rights, that are more akin to a buyout or Royalty free licensing.
Such “rights-grabbing” should be in the forefront of any rights holders mind when negotiating terms, to ensure that you are left being able to utilise, monetise, control & protect the rights in your own creative works.
Defining a license by the specific product/publication, issue/edition, print-run/duration & territory is the best way to limit the rights granted to what’s actually required for publication.
Things to be mindful of:
This goes hand-in-hand with –
It is wise to always keep in mind what are the minimum rights actually required for the product/publication to be produced. Anything additional, ask yourself “what is that for?” – if it’s not explained in the contract, ask for explanation and if you are happy to grant such rights as they seem necessary for the agreement, then make sure they are defined in writing within the agreement (see “Entire Agreement” above).
Trevor Jennis McDermott, GAP Photos
www.gapphotos.com
