Building Sovereign European Infrastructure for Content Trust : An Interview with Xavier Bruni

Building Sovereign European Infrastructure for Content Trust — Interview with Xavier Bruni | Provenance For Trust
September 2026 Interview Digital Sovereignty Content Provenance

Building Sovereign European Infrastructure for Content Trust: An Interview with Xavier Bruni

A month after entering beta, ProvenanceForTrust — the Provenance program's platform — lets media outlets and information professionals verify the origin of content and document its provenance. It's built as sovereign European infrastructure, aligned from the ground up with the AI Act, the DSA, and the EMFA. Program lead Xavier Bruni reviews the first month.

Credit: CEPIC

Xavier Bruni, Provenance for Trust program lead


Photo Credit: CEPIC
🇪🇺
European program
A six-member European consortium
Launched January 2025, targeted for completion by the end of 2026
57%
overall program progress before summer 2026
400
Journalism Trust Initiative-certified outlets offered launch access
6
partner organizations in the European consortium

What is the Provenance program, and what problem is it solving?

Provenance is a two-year program, launched in January 2025, built to answer a very concrete question for media outlets and information professionals across Europe: how do you verify where a piece of content came from, and how do you document its provenance, at a time when AI-generated or AI-modified content is everywhere?

It isn't a standalone research exercise. It combines product development with sustained work on aligning the platform with the European regulatory framework. The platform now live at app.provenance4trust.org is the first concrete output of that effort.

What can the platform do today?

It's built around two components.

Component 01
Verification
Analyzes a piece of content — image, audio, or video — for signs it may have been generated or modified by AI, using a set of baseline forensic tools.
Component 02
Labeling
Lets a media outlet, agency, or content creator document the provenance of what it publishes: identity, source, and AI involvement where relevant — through C2PA manifests, metadata, and watermarking.

Both components currently cover image, audio, and video. Text processing isn't available yet — we address that below.

Who is building ProvenanceForTrust?

ProvenanceForTrust is a genuine European consortium effort, not a single-vendor product. This mix of expertise lets the consortium advance the product and its legal foundations at the same time — a deliberate choice to build infrastructure that's accountable to European rules by design, not retrofitted to them.

Coordination
L'Atelier

Leads the consortium, coordinates the program, and contributes to platform development.

Detection
UncovAI

Develops the AI-generated content detection engine.

Labeling
TrustMyContent

Provides the provenance-labeling layer via the C2PA standard and watermarking, with partner IMATAG.

Certification
Reporters Without Borders — JTI

Created and runs the Journalism Trust Initiative, the certification framework for transparency and integrity in media.

Research
Sciences Po Médialab

Studies the platform's impact on trust between media outlets and their audiences.

Regulation
CEPIC

Supports alignment with the European regulatory framework — the AI Act, the DSA, and the EMFA.

One month after the beta launch: what's the early assessment?

Launch offers (free access) went out to the 400 media outlets certified under the Journalism Trust Initiative, as well as to all CEPIC members. Sign-ups are trickling in — still more cautiously than the team would like — with a stronger push expected as activity resumes in September.

The beta launch was deliberately designed as a learning step rather than a finished product. The feedback gathered so far is already shaping the next phase of development.

What has the early user feedback revealed?

The team held initial conversations with media outlets, fact-checkers, IT managers, and photographers from around the world — France, the US, Indonesia, the Czech Republic, Spain, Latvia, Mongolia, and more. Needs and expectations vary significantly by profile, but one need is universal: protecting digital content.

The most anticipated feedback, unsurprisingly, concerns text. The original roadmap planned text detection alongside images from the first phase; that isn't in place yet, and the question is still being worked through internally.

"It's something we address directly, particularly because it's sensitive for news media in light of Article 50 of the AI Act, which governs disclosure requirements for certain AI-generated texts used to inform the public on matters of public interest."

Xavier Bruni, on text detection ahead

Where does platform development stand?

Before summer 2026, the program's overall progress stood at 57%, with full program completion targeted for the end of 2026. The technical foundation is complete, the beta platform is being finalized, and an advanced version is in development.

Identity layer (CAWG) — under development.
Video and deepfake detection — ongoing improvement.
Hardening detection models.
Video and audio watermarking — planned extension.

ProvenanceForTrust is meant to be read as an evolving platform, not a finished product.

How central is the regulatory question — AI Act, DSA, EMFA — to this work?

Central, but not the whole story: the core of the program is provenance and traceability, technical subjects first and foremost. As regulatory obligations accumulate across multiple frameworks, though, this dimension keeps growing in importance — it's a workstream in its own right, not an afterthought. CEPIC works specifically on analyzing the AI Act, the DSA, and the EMFA to identify the principles that should guide the platform's development.

One point is important to state clearly: the verification and labeling features on offer are aids to verification and traceability. On their own, they don't guarantee a piece of content's authenticity, nor that a given use complies with the legal obligations applicable to each actor. That analysis remains the responsibility of users — the platform's role is to give them information that's as clear and accurate as possible to help them get there.

The goal hasn't changed

To give media outlets and information professionals reliable, sovereign, European tools to document and verify the provenance of what they publish.

-
Xavier Bruni, Provenance for Trust program lead

Photo Credit: CEPIC

What's next for ProvenanceForTrust?

Continued work on the advanced version of the platform, extending detection to text, and refining integration into real editorial workflows — notably through APIs designed for content management systems, currently being built by L'Atelier and TrustMyContent.

The objective hasn't changed since day one: give media outlets and information professionals reliable, sovereign, European tools to document and verify the provenance of what they publish, within a framework that stays rigorously aligned with regulatory requirements.

Frequently asked questions

What is Provenance For Trust?
A two-year European program, launched in January 2025 by a six-member consortium, developing a platform to verify and document the provenance of image, audio, and video content as AI-generated content becomes more widespread.
Does the platform detect AI-generated text?
Not yet. Text detection is on the roadmap but isn't available in the current beta; the question is still being worked through, particularly in light of Article 50 of the AI Act on disclosure of AI-generated text used to inform the public.
Who can use the beta platform?
Media outlets, fact-checkers, agencies, and content creators. Free access was offered to the 400 Journalism Trust Initiative-certified outlets and CEPIC members; the platform remains open in beta at app.provenance4trust.org.
Does the platform guarantee its users' regulatory compliance?
No. The verification and labeling tools are aids to verification and traceability; on their own, they don't guarantee a piece of content's authenticity or compliance with the legal obligations applicable to each actor, which remains the user's responsibility.

Are you a journalist, fact-checker, or IT lead at a media organization?

The platform is open in beta. Try it out and share your feedback — it directly shapes what we build next.

Try the beta →

Have a question about the program or want to talk to the team?

Contact us →
Interview Digital Sovereignty Content Provenance AI Act DSA EMFA C2PA ProvenanceForTrust
Précédent
Précédent

Construire une infrastructure souveraine européenne pour la confiance dans les contenus : entretien avec Xavier Bruni

Suivant
Suivant

IA et présidentielle française de 2027 : le vide réglementaire que les journalistes doivent surveiller