Is AI training data protected by copyright?

Yes. Copyright attaches automatically at creation under the Berne Convention (Art. 5) in all 181 member countries. Scraping copyrighted works to build a training dataset creates copies of protected material. Whether it constitutes infringement depends on jurisdiction and whether applicable exceptions, such as the EU CDSM Directive Art. 4 TDM exception, apply and have not been opted out of by the rights holder.

Can I use scraped web data to train an AI model?

Only if the source material is explicitly licensed for that purpose or the applicable TDM exception covers it. In the EU, rights holders may opt out of TDM exceptions via machine-readable signals such as robots.txt. In Switzerland, no general TDM exception exists. Scraping for commercial AI training without a licence is legally risky. Always audit your dataset sources and document licensing status before training.

What does the EU AI Act require about training data documentation?

Under EU AI Act Article 53, providers of general-purpose AI models must maintain technical documentation of training data sources, keep a sufficiently detailed public summary, and comply with EU copyright law including TDM opt-outs. For high-capability models above the 10^25 FLOPs threshold, additional adversarial testing and incident-reporting obligations apply. Non-compliance exposes providers to fines up to 3% of global turnover.

How can a cryptographic timestamp prove training data provenance?

A qualified electronic timestamp under eIDAS or ZertES embeds a cryptographic hash of your dataset at a specific point in time, anchored to a trusted time source. This creates an immutable, court-admissible record proving the dataset existed in that exact form on that date. It is the primary technical mechanism for demonstrating legitimate, pre-existing sourcing when rights holders or regulators audit your data pipeline.

Which jurisdictions have the strictest AI training data copyright rules?

The EU has the most comprehensive framework, combining CDSM Directive TDM exceptions with opt-out rights and the EU AI Act documentation requirements. Switzerland has no TDM exception, making unlicensed scraping riskier than in the EU. The US applies fair use analysis, which is less predictable. Japan has the broadest TDM exception, permitting scraping for AI training even of commercial works without opt-out mechanisms.

How do I protect my content from AI scraping under EU law?

Under EU CDSM Directive Article 4(3), rights holders may opt out of text and data mining exceptions by publishing a machine-readable reservation, such as X-Robots-Tag: noai or a robots.txt disallow directive. To prove your content existed before an AI model was trained, obtain a qualified electronic timestamp under eIDAS Article 41(2). A qualified timestamp from a QTSP like Swisscom Trust Services creates a legal presumption of accuracy that is admissible across all 27 EU member states. If an AI model reproduces your timestamped content, the timestamp becomes your primary exhibit in an infringement claim. It cannot be backdated and is cryptographically verifiable.

Protection des données d'entraînement IA par le droit d'auteur : ce que les développeurs doivent savoir en 2026

AI Technology

Retour au Blog

Share Share

Protection des données d'entraînement IA par le droit d'auteur : ce que les développeurs doivent savoir en 2026

Les données d'entraînement IA sont soumises au droit d'auteur — le contenu web extrait par scraping n'entre pas automatiquement dans le domaine public. Découvrez ce que la conformité à l'IA Act européen, la concession de licences et les horodatages eIDAS signifient pour les développeurs en 2026.

Swiss Trust Layer Editorial Team· Legal Content

·June 12, 2026·Last updated June 12, 2026· 8 min de lecture

En octobre 2024, une start-up berlinoise specialisee dans l'IA a clos une Serie A et entame une procedure de due diligence avec un client entreprise de premier rang. L'equipe juridique du client a pose une seule question: quand chaque jeu de donnees a-t-il ete constitue, et pouvez-vous prouver que cette date est anterieure a votre date de conformite au mecanisme d'exclusion? Le directeur technique de la start-up ne disposait d'aucun horodatage qualifie sur aucun jeu de donnees. Deux semaines plus tard, la transaction etait bloquee. Les memes jeux de donnees ont ensuite ete contestes par un editeur de presse dont l'exclusion robots.txt etait active depuis mars 2024.

Alors que l'application du reglement europeen sur l'IA commence en 2026, la documentation de la provenance des donnees est une obligation legale pour les fournisseurs de modeles d'IA a usage general. L'absence d'horodatages sur les jeux de donnees d'entrainement est la lacune la plus frequemment constatee lors des inspections reglementaires.

Consultez le guide complet sur la propriete des contenus IA

Protégez votre travail avec Swiss Trust Layer AG

Scellez votre propriété intellectuelle avec un e-Sceau prouvé en justice, soutenu par Swisscom Trust Services.

Réserver une Démo Gratuite

AI & Technology

If someone asks you to prove your content is compliant, what do you actually send?

When a client, platform, or regulator asks for proof of AI content compliance, a disclosure label is not enough on its own. This is the evidence pack that actually holds up: a timestamp, a signature, a content hash, and a link anyone can check without contacting you.

August 10, 2026Read more →

AI & Technology

Week in review: Article 50, copyright proof, and what changed

Article 50 took effect on 2 August. A week of writing about it comes down to one idea: the rule asks you to make a statement, and a statement is only as good as what you can produce when somebody asks you to back it up.

August 8, 2026Read more →

AI & Technology

The four AI labels every EU publisher needs, and what each one means

AI generated, AI modified, AI assisted, human authored. Four statements cover almost everything a publishing team produces. This is what separates them, how to decide in under a minute, and why an inaccurate label is a worse position than a missing one.

August 4, 2026Read more →

AI & Technology

A checkbox is not proof: labelling versus evidence under the AI Act

A label and a record look identical on a page and behave completely differently the moment somebody questions them. This is what separates a self-declaration from evidence, why the usual internal records do not close the gap, and what a qualified timestamp changes.

August 3, 2026Read more →

AI & Technology

Article 50 is live today. Here is how to prove your content is compliant.

The transparency obligations in Article 50 of the EU AI Act apply from today. This is what the obligation actually requires, why a label alone leaves the burden of proof with the publisher, and the four parts of a record that somebody outside your organisation can check.

August 2, 2026Read more →

Protection des données d'entraînement IA par le droit d'auteur : ce que les développeurs doivent savoir en 2026

Related Articles