AI content authenticity and provenance
Prove the origin of everything your AI generates.
From 2 August 2026, the EU AI Act requires AI-generated content to be marked, machine-readable, and disclosed. Filigrane gets you there two ways: Filigrane Origin, our provenance API and SDKs now in development, and implementation by Webisoft, where our engineers build watermarking, C2PA Content Credentials, and provenance into your product today.
EU AI Act Article 50 transparency obligations apply in
Transparency obligations under Article 50 apply from 2 August 2026. Generative AI systems already on the market have until 2 December 2026 to bring marking and detection into line. Penalties reach EUR 15M or 3% of global annual turnover.
Two ways to get there
A product on the way, a team ready now
The same provenance engineering, delivered how your organization needs it: as an API you integrate, or as a system we build for you.
The product
Filigrane Origin
The provenance engine as an API and SDKs: mark, sign, verify, and evidence AI content from one integration. In active development, waitlist open now.
Join the waitlistThe services
Implementation by Webisoft
Our engineers build watermarking, C2PA Content Credentials, and provenance ledgers into your product, or deliver the whole stack self-hosted in your infrastructure. Available now.
Scope an implementationThe knowledge base
Five obligations your product now carries
Engineering-grade, primary-source guidance across the full content authenticity stack: the legal duty, the marking technology, the credentials standard, deepfake disclosure, and document provenance.
EU AI Act Article 50: every obligation, date, and penalty, explained
EU AI Act Article 50 requires watermarking and disclosure of AI content from 2 August 2026. Full obligations, timeline, and penalties, sourced and dated.
WatermarkingAI content watermarking: the engineering view
How AI content watermarking embeds detectable signals in text, images, audio, and video at generation time, and how to build or buy a pipeline.
Content CredentialsC2PA and Content Credentials, explained
What C2PA and Content Credentials are, how manifests, signing, and verification work, and who backs the standard, explained for product teams.
Deepfake provenanceDeepfake provenance: proving media is real instead of chasing fakes
Why cryptographically signed provenance beats deepfake detection as a long-term trust strategy, how C2PA signing works, and where its limits are.
Document provenanceDocument Provenance: How to Prove a Document Is Authentic and Unaltered
What document provenance means, how hashing, signatures, timestamps and audit trails prove authenticity, and when simpler tools are enough.
Why Filigrane
Built by the engineers who implement it
Most Article 50 coverage stops at "you must comply." We are the engineers who wire watermarking, C2PA Content Credentials signing, and provenance verification into production systems. Every page on this site is written from that build experience, cited to primary sources, and maintained as the standards evolve.
Standards-neutral
We assess SynthID, C2PA, and open watermarking on their merits, then specify what fits your stack, your media types, and your risk profile.
Implementation-first
Not policy memos. A reference architecture end to end: embed at generation, disclose in the interface, verify downstream, and keep evidence your auditors can use.
Built by Webisoft
A Montréal software engineering firm. We scope the system, build it into your product, and hand your team something that ships. SOC 2 available.
Assess your Article 50 readiness
A confidential two minute self-assessment. See where your product stands under Article 50 and what it takes to close the gap.
Start the assessment