RealityFoundation
Certify your media at capture — tamper-proof, blockchain-anchored, C2PA-compliant.
2025
Founded
4
Research themes
3
Founding members
🇪🇺
EU
Based in Europe
The Problem
Today, there is no reliable way to prove with certainty whether a photo, video, or document was AI-generated. Detection-based approaches are fundamentally reactive — they lag behind generation models that keep getting better at producing realistic, harder-to-detect synthetic content.
Journalism, legal evidence, insurance claims, and public accountability all depend on one assumption: that what we see is real. When that assumption breaks, truth becomes negotiable. Bad actors fabricate freely, and the honest lose the ability to prove they're telling the truth.
This is a closing window. Once AI-generated content becomes indistinguishable from reality, retroactive verification becomes impossible — any unverified record from the past stays permanently disputable. Trust also collapses asymmetrically: once people stop believing digital evidence, restoring that trust is extraordinarily difficult. The authenticity layer has to exist before scepticism becomes the default.
Aerial photo of a snowstorm over a city, widely shared online
Real — and disputed
A viral aerial photo of a snowstorm was dismissed by many as AI-generated. It wasn't.
Source: BBC News
Side-by-side comparison of a real photo and an AI-generated image purporting to document the same event
Fake — and convincing
AI-generated images purporting to document a conflict zone circulated widely before being debunked.
Source: Financial Times
Our Approach
We're building the infrastructure to certify media at the moment of capture — before manipulation can occur. Raw media and its metadata (timestamp, GPS, device fingerprint) are hashed instantly. That hash is signed and anchored to a public blockchain, producing a permanent, tamper-proof certificate of authenticity. The result is natively compatible with C2PA, the open content-provenance standard backed by Adobe, Google, Microsoft, OpenAI, and Amazon.
How it works
A high-level look at the certification process. An interactive demo is coming with our MVP.
Step 01
Raw media and its metadata — timestamp, GPS, device fingerprint — are collected the instant the shutter clicks.
Step 02
A cryptographic hash of the capture is generated instantly — unique to that exact file and its metadata.
Step 03
The hash is signed and written to a public blockchain, with the underlying content stored on IPFS.
Step 04
The result is a permanent, tamper-proof certificate of authenticity — natively compatible with C2PA.
Four interlocking areas that together address the challenge of maintaining verifiable reality.
Defining what cryptographic and physical evidence constitutes proof that a digital asset was captured, not generated.
Evaluating blockchain-based timestamp mechanisms for evidentiary integrity across legal and journalistic contexts.
Studying how manipulated imagery enters insurance claims processes and how provenance records can prevent it.
Developing open methodologies for detecting and attributing synthetic media at a forensic standard of rigor.
The EU AI Act, the Digital Services Act, and the forthcoming Media Authenticity Regulation are creating binding obligations around synthetic media transparency. RealityFoundation is positioned to provide the independent academic grounding these frameworks require — from methodology to evidence standards.
Requires watermarking of AI-generated content. Our research informs verification methodology.
Discover more →
Mandates risk assessments for synthetic media on platforms. We provide independent benchmarks.
Discover more →
We study interoperability between blockchain records and emerging open standards like C2PA and IPTC.
Discover more →
Get Involved
We are building a multidisciplinary research community. There is a role for every background.