Opinion

Signed, filed,
never read again.

The signature was the easy part. Almost every release dispute we hear about turns on something else entirely: a territory nobody noticed, a medium the form never covered, a use that did not exist when it was drafted. And it surfaces years later, with someone you got on with perfectly well.

ProntoID  ·  Brooks & Keitt Sàrl · · 12 min read

A photographer we spoke to had been shooting with the same handful of models for nearly a decade. Good relationships, repeat work, everybody paid on time. When we asked what his releases said about video, he had to go and look. They said nothing about video. He had been shooting video for four years.

Nothing bad happened to him. That is rather the point — nothing bad happens for a very long time, and then the circumstances change and the paperwork is asked to do a job it was never drafted for.

Trust is not the failure

There is a version of this argument that treats photographers as careless and models as litigious. It is wrong and it is insulting, and it also misdescribes how these disputes actually start.

They rarely start with a fight. They start with a change of circumstances. Someone leaves the industry and takes a job where the content is a problem. Someone gets married, or divorced, or has a child, or a family member finds something. Someone moves to a country with strong image rights and gets advice they did not previously have. An agency does an audit. A partner objects.

At that moment the relationship stops being the operative fact. Everyone is perfectly civil, and the question quietly becomes: what does the document say? If nobody has read it since the day it was signed — and almost nobody has — then the answer is a genuine surprise to both sides.

A good professional relationship protects you right up until the moment it is no longer the thing being relied on.

Six quiet failures

These are not exotic. Every one of them comes from a form that was signed willingly by someone who meant it, and every one of them is invisible until somebody goes looking.

Territory

Someone adapted a template that excluded a market, because the original drafter had a reason to. Three years later a platform syndicates worldwide and nobody remembers the carve-out is there. Germany, France and Italy are the usual suspects, because those are the markets template authors most often single out.

Medium

The single most common failure we see. A form written for stills, used for video — and forms written for stills very often do not mention moving image at all. Then behind-the-scenes footage, a reel, a clip. Different right, same folder, no cover.

Purpose

“Portfolio and promotional use” is not commercial licensing. It is a phrase people sign happily because it sounds modest, and it is precisely why it will not carry a stock sale, a paid subscription tier, or an advertising placement.

Term

Some releases expire. A five-year licence signed in 2019 is a problem in 2026, and the content did not stop being published on the anniversary. Nobody diaries these, because the form goes in a folder the moment it is signed.

Identity and naming

Agreeing that an image may be published is not agreeing that a legal name may appear beside it. Performers who work under a stage name often assume the distinction is obvious. It is obvious to them. It is frequently absent from the paperwork.

Silence about the future

A release from 2015 says nothing about training a model on the images, because nobody was asking. Silence is not permission — but it is an argument, and it is an argument you will be having with someone who did not agree to it and never imagined it.

Notice what these have in common. Not one of them is a forgery, a coercion, or a minor. They are all scope — the boundaries of a permission that was genuinely given. And scope is the part nobody negotiates, because on the day of the shoot everyone is thinking about the shoot.

The German example

Territory carve-outs feel abstract until you look at a jurisdiction that takes image rights seriously.

Germany has protected the Recht am eigenen Bild — the right to one's own image — since 1907. Section 22 of the Kunsturhebergesetz provides that images of a person may only be disseminated or publicly displayed with that person's consent, with a narrow set of exceptions in section 23 covering things like figures of contemporary history, incidental appearances and gatherings. How the KUG sits alongside the GDPR has been argued over since 2018 and is not fully settled, but German courts have continued to apply it.

Here is the part that usually surprises people, and it cuts against the way this argument is normally made. German courts have treated consent under section 22 as relatively difficult to revoke — only for special reasons, such as a material change in the circumstances in which it was given. At least one appellate court has held that Article 7(3) GDPR does not change that result, on the basis that the KUG operates as the more specific regime.

So the German risk is generally not that a model wakes up and withdraws. It is that content is disseminated outside the scope that was consented to — and Germany has a long, well-developed practice of injunctive relief for exactly that. A release that excludes Germany, or a use the release never covered, is not a technicality there. It is the whole case.

France, Italy and Spain each have their own version of this. The details differ; the shape does not. The further your content travels, the more legal systems get an opinion about what you were permitted to do with it — and the less likely it is that a single page signed in a studio anticipated all of them.

Two different questions

This is the conflation that causes the most trouble, and it is worth stating plainly because a lot of people in this industry have it backwards.

§ 2257 asks: was this person an adult, and can you prove it? It is a federal criminal record-keeping regime about age and identity. It does not care what the performer agreed to.

The release asks: did they agree to this use? It is a civil question about permission, and it is governed by whatever law reaches the publication — which may be German, French, or somewhere neither of you has been.

You can be immaculate on the first and lose badly on the second. A perfect § 2257 file proves the performer was 24 and that you examined their passport. It says nothing whatsoever about whether they agreed to a video being distributed in a market their release excluded. If your compliance thinking is entirely organised around records retention, you have covered one of two exposures and may not have noticed the other exists.

Why paper cannot keep up

The underlying problem is a mismatch of shape. A paper release is a static artifact: signed once, describing a set of permissions in prose, filed. Distribution is dynamic: content gets re-cut, re-published, syndicated, licensed onward, embedded, and eventually handled by people who have never met you and have certainly never seen the release.

A platform that receives your content cannot read the PDF in your filing cabinet. Even if you send it, nobody is going to parse a paragraph of legal prose per asset at upload time and correctly extract that video is excluded and Germany is out. So the boundaries you carefully agreed simply do not travel with the file. They stay where they were written, and everything downstream operates as though permission were unlimited.

The permission is specific. The file is not. Everything after that is somebody assuming.

What a better record looks like

The fix is not a better-worded release. It is making the boundaries structured instead of narrative, and attaching them to the specific content rather than to a person or a shoot.

That is what ProntoTag does. Permissions are recorded as discrete states rather than paragraphs: each one granted, prohibited, or unspecified. Both parties are identity-verified, the model approves the specific content rather than a general arrangement, and the result is a record a platform can actually read.

Two design decisions matter more than the rest:

  • Unspecified is not permission. A boundary nobody addressed reads as unaddressed, not as a yes. This is the direct answer to the 2015 release that is silent about AI training.
  • Some permissions are closed by default. Model training, biometric enrolment, sublicensing and publication of a legal name all start prohibited and have to be opened deliberately. These are the four where an assumption in the wrong direction is hardest to undo.

And because the record is per depiction, the question “may we publish this clip in Germany” has an answer that a machine can retrieve, rather than an answer that depends on someone remembering a conversation from 2019.

What we cannot do

Now the part that most vendors would leave out, and that you should insist on hearing before you buy anything in this category.

ProntoTag cannot enforce anything. It is an information layer. It records what was permitted, by whom, for what content, in a form that can be read and acted upon. It cannot reach into a distribution pipeline and stop a file moving, and it cannot compel a platform to check the boundaries before publishing.

If a platform never integrates the permissions, what you have is an excellent record of a limit that was exceeded. That is genuinely valuable when a dispute arrives — it is the difference between arguing about what was agreed and pointing at it — but it is evidence, not prevention, and the two are not the same thing.

Which is why the platform side of this matters as much as the photographer side. A platform that consumes the permission data can gate publication on it: refuse the German territory, refuse the video format, flag the record whose term has expired. That is prevention. It requires the platform to do the work, and no amount of good record-keeping upstream substitutes for it.

So when we say ProntoTag is the safest option available to a photographer, we mean precisely that and no more: it is the strongest position you can occupy. Whether that position is honoured downstream depends on people who are not you. We would rather say so now than have you discover it at the worst moment.

Where to start

Nothing here requires a project. It requires an afternoon and a slightly uncomfortable hour.

  • Pull five releases at random and read the scope clauses — not the signature block. Territory, medium, purpose, term. Compare them against what you are actually doing with that content today.
  • Expect to find at least one mismatch. Most people do. Finding it is the good outcome.
  • Where the performer is still reachable, a fresh per-depiction record is the clean fix, and the conversation is far easier now than after a letter arrives.
  • Where they are not, at least know where the exposure sits. Our archive exists for exactly that case, and it is honest about being the weaker option.
  • If you buy content from others, the same scope questions apply to paperwork you did not draft — see buying content and § 2257.

There is a working checklist covering this alongside the record-keeping and data protection side, with a track for photographers.

This article is opinion and general information, not legal advice. ProntoID and Brooks & Keitt Sàrl are not a law firm. Image rights, personality rights and release enforceability vary substantially between jurisdictions, and the German position described above is a summary of a contested area. Take qualified advice on your own arrangements.

Frequently Asked Questions

The awkward ones

We have a good relationship. Is this really a risk?

The relationship is almost never what fails. What fails is a change of circumstances years later — a career move, a new partner, a family finding out, a country with strong image rights. At that point the only thing standing between you is a piece of paper neither of you has read since the day it was signed, and the question stops being whether you trusted each other and becomes what the document actually says.

What is the most common scope error?

Medium. A release drafted for stills, used for video. The two are different rights and a form written for one frequently does not mention the other at all. Territory is second, usually because someone adapted a template that carved out a market they never intended to publish in anyway — until a platform syndicated the content worldwide.

Does § 2257 compliance protect me from a release dispute?

No, and conflating the two is common. § 2257 asks whether the performer was an adult and whether you can prove it. The release asks whether they agreed to this particular use. You can be immaculate on the first and lose badly on the second. They are separate obligations answering separate questions, and satisfying one says nothing about the other.

What about uses that did not exist when the release was signed?

A release from 2015 says nothing about AI training, because nobody was thinking about it. Silence is not permission. If a use is not covered, the safe reading is that it is not permitted — which is why a modern consent record should treat new uses as closed by default rather than leaving them to be argued about later.

Can a model withdraw consent in Germany?

Less easily than many people assume, which cuts both ways. German courts have treated consent under § 22 KUG as revocable only for special reasons such as a material change in circumstances, and at least one appellate court has held that GDPR Article 7(3) does not change that. So the German risk is generally not sudden revocation — it is publishing outside the scope that was agreed, which is actionable and where injunctive practice is well established.

Does ProntoTag stop content being distributed beyond its permissions?

No, and we want to be unambiguous about that. ProntoTag records what was permitted, by whom, for what content, in a form a platform can read and act on. It is an information layer. It cannot reach into a distribution pipeline and stop a file moving. If a platform ignores the boundaries or never integrates them, what you have is an excellent record of a limit that was exceeded — valuable in a dispute, useless as prevention.

Then what is the point of structured permissions?

Three things. Nobody can later claim the boundary was unclear. The platform has no excuse for not knowing, which changes where responsibility sits when something goes wrong. And on integrated platforms the boundaries can actually gate publication, which is prevention rather than evidence. The first two are worth having even where the third is not available.

What should we do about releases we already hold?

Read a sample of them properly — not the signature block, the scope clauses — against what you are currently doing with the content. Most people find at least one mismatch. Where the performer can still be reached, a fresh per-depiction record is the clean fix. Where they cannot, at least know where the exposure is rather than discovering it in a letter.

Go and read five of them

Then, if you find what most people find, there is a better way to record the next one.

See ProntoTag Readiness checklist

Brooks & Keitt Sàrl  ·  Place du Midi 30, 1950 Sion, Switzerland