AI crawlers strip metadata in seconds. Sealing an image at export creates a dated, globally recognized record proving you made it, before it turns up in someone else's feed or training set.
Your image leaves your camera roll once. After that it moves through a phone, a laptop, a content calendar, an Instagram post, a client email thread, and eventually the open web, where an AI crawler can pick it up in seconds. Once it is scraped, screenshotted, reposted, or pulled into a training set, the file that reaches other people rarely still has your name attached to it. EXIF data gets stripped on upload. Compression removes metadata. A repost strips the caption. What is left is a picture that looks like nobody made it.
For photographers and creators, that is not a hypothetical problem anymore. Stock libraries, dataset scrapers, and social aggregators pull images at scale, and a single viral repost can put your work in front of millions of people who have no idea it is yours. If a dispute ever comes up, whether it is a stolen client photo, an AI company that trained on your portfolio, or a competitor lifting your work for their own feed, the question is always the same: can you prove, with a specific date, that the file existed and that you made it.
Under the Berne Convention, copyright protection attaches the moment a work is created, in every one of the treaty's member countries, with no registration required. That is good news in principle. In practice, automatic protection only helps you if you can show, later, exactly when a specific file existed and that it came from you. A JPEG or RAW file with no metadata and no paper trail does not do that on its own. Courts, platforms, and licensing bodies want evidence, not a general legal principle.
This is the gap creators keep running into. The law already protects your work. What is missing is a record, made at the moment of creation, that stands up when someone else claims the image or when an AI system has already trained on it and there is nobody left to ask.
The fix is not complicated, but the timing matters. Instead of uploading a finished image and hoping nobody strips the metadata, seal the file the moment you export it, before it goes anywhere. A qualified electronic timestamp, as defined under the EU's eIDAS Regulation, creates a cryptographic record of a specific file existing at a specific moment, issued by a regulated trust service provider. That record does not depend on the platform you post to, and it survives the image being copied, cropped, screenshotted, or reposted a hundred times over, because the seal is tied to the original file, not to whatever caption or metadata happens to travel with a copy of it.
This is the evidentiary base layer under everything else: proof of possession of a specific file at a specific time. It answers the first question anyone will ask you if your work turns up somewhere you did not put it.
Here is the part that matters most for anyone whose work moves internationally, which is nearly every working creator today. A dispute over a scraped image rarely stays inside one jurisdiction. Your photo might be shot in Zurich, licensed to a client in New York, and scraped by a training pipeline that could be running anywhere. If your only proof of authorship is tied to a single country's registration system, it can leave you stuck arguing jurisdiction before you even get to argue the facts.
A sealed record built on a Swiss qualified timestamp works the other way around. It is built to be globally recognized proof, not a claim that only holds up locally and has to be re-established every time it crosses a border. That asymmetry, proof that travels with the file instead of proof that stops at a border, is the difference between a record you can actually use and one that just sits in a folder looking official.
A timestamp proves you had the file. It does not, on its own, help you find out if someone else is using your content, or if an AI system trained on it. That is a separate problem, and it needs a separate piece: a content hash, generated from the image itself and registered into a public, searchable database, following the ISCC (International Standard Content Code) approach. The hash lets your content be matched and tracked for infringement or AI-training licensing purposes without ever exposing the original file to anyone doing the search. Nobody can reverse a hash into your photo. They can only confirm a match against it.
For creators specifically, this is the piece that starts to matter as AI companies face more scrutiny over what they trained on. The EU AI Act requires providers of general-purpose AI models to keep, and in some cases publish, a sufficiently detailed summary of the training data they used. Separately, the EU's Copyright in the Digital Single Market Directive gives rightholders a way to reserve their works from text and data mining under Article 4, provided the reservation is expressed in a machine-readable way. A registered content hash is exactly the kind of machine-readable signal that framework is built around. Sealing the file proves you made it. Hashing the content is what makes it findable later, on your terms.
None of this is worth much if the proof cannot be tied to an actual, verified human, and if you lose access to your own record the moment you lose a password. Identity verification (KYC) ties the seal and the hash to a specific real person, not an anonymous account or a bot, which matters if you ever need to assert the claim in front of a court or a platform's takedown team. And because losing a device or a credential should not mean losing years of proof, a proper setup includes an escrow recovery path: your identity can be re-verified in person by a local, KYC'd agent even if every digital credential is gone, so the record survives you losing access to it.
A marketplace or stock site lists your photo under someone else's name. Without a sealed, dated record made before the listing appeared, you are arguing your word against theirs.
A dataset audit, your own or a journalist's, surfaces your work inside an AI training corpus you never agreed to. A registered content hash is what lets you demonstrate the match without handing over your entire portfolio to prove the point.
A client dispute breaks out over who owns the final delivered files after a shoot. A timestamp made at export, before delivery, settles the timeline before it becomes a he-said-she-said conversation.
For a working photographer or creator, the workflow is simple to describe even if the underlying cryptography is not.
This will not stop a scraper from taking your image. Nothing fully does, and anyone promising otherwise is not being straight with you. What it does is put you in a position where, if your work turns up in a place you never put it, you are not starting from zero. You have a timestamped, globally recognized record that the file existed, and a searchable hash that connects it back to you without ever handing over the original.
Swiss Trust Layer is built for creators and photographers who need exactly this: a sealed, verifiable record of a file's existence, tied to a real identity, made the moment the work is finished rather than after it has already been copied a hundred times. If your portfolio, your client deliverables, or your published work has ever ended up somewhere you did not put it, start sealing new work before it leaves your hands and see how the record holds up.
Protect your work with Swiss Trust Layer AG
Seal your intellectual property with a court-proof e-Seal backed by Swisscom Trust Services.
Book a Free Demo
September, in one checklist: what's live, what's changing, what to fix first
This week closed with a record AI copyright settlement, a new EU disclosure law in force, and a jury trial underway that could reshape how image generators are treated. Here is a practical checklist covering what to check and fix in September, from AI content disclosure to dated proof of your own work.
The AI Copyright Lawsuit Count Nearly Tripled in a Year. Here's What Actually Changed
AI-related copyright lawsuits in the US grew from roughly 30 to over 70 in 2025, according to the Copyright Alliance. Courts are now ruling instead of settling quietly, and the EU has introduced its own transparency rules for AI-generated content. Here is what is driving the surge and what it means for anyone publishing AI-generated material.
A Five-Minute Habit That Changes the Math Before Any Dispute Starts
Proof made after a dispute starts always looks assembled for the occasion. This post walks through what a qualified timestamp and hash actually capture, why identity has to be attached to make a claim yours, and why sealing a file is a five-minute habit, not a project for later.
Fair Use for Training, Not for Storage: The Line Courts Are Actually Drawing
In Bartz v. Anthropic, a federal court found that training an AI model on copyrighted books could plausibly be fair use, while separately storing pirated copies of those books was not. That distinction, not a blanket ruling on AI training, drove the 1.5 billion dollar settlement approved in July 2026. Here is what the training versus storage line actually means for legal and compliance teams.
What a $1.5 Billion Copyright Settlement Actually Tells You About Proof
A $1.5 billion settlement did not happen because one side was more right. It happened because the facts, which works, whose files, when they were accessed, could be proven at scale. That distinction applies to anyone who has ever needed to show a piece of work was theirs first.
