Manifesto for digital trust.

The reality

For two centuries, seeing was believing. Photography, video, and audio recordings built a shared consensus on what was real. In 2025, that consensus collapsed. AI-generated content surpassed human-produced content in volume. Deepfakes are no longer a lab demo, they are commodity goods, produced in minutes, indistinguishable to the eye and ear.

We have lost the war for reality. Not by decision, by default.

Why current approaches fail

Three families of tools try to answer this crisis. All miss the mark.

Automatic detection of AI content is a race defenders cannot win. Every gain in detectors triggers a matching gain in generators. The asymmetry structurally favors the attacker.

Provenance standards like C2PA, backed by Adobe, Microsoft, BBC, and Sony, do excellent technical work on file traceability from the creation tool. But the signer's identity is optional. You can sign provenance without engaging verified human responsibility. Necessary, not sufficient.

AI watermarks like SynthID flag content as AI-generated. Useful, but limited to AI content, circumventable, and silent on the human responsibility behind distribution.

The conceptual shift

We propose to invert the burden of proof.

Presumption of fiction by default. Verified identity by default. Any unsigned digital content should be treated as potentially fictional until proven otherwise. Any signed content engages a verified human identity, attested by a trusted third party, on a tamper-proof public registry.

This inversion changes the question. We stop asking "is this content true?", a question pixel analysis can no longer answer. We ask "who takes responsibility for it?", a question a cryptographic signature linked to a verified identity answers without ambiguity.

The protocol

The Alethea protocol fits in one sentence. A verified identity and a piece of digital content are linked by a cryptographic signature, and that signed link is recorded on a public blockchain.

Two entities. A signature linking them. An archive layer guaranteeing immutability.

1. ENTITY 01 · WHO

   Verified identity.

   Verification compliant with a recognized international standard: eIDAS substantial or high, NIST 800-63 IAL2 or IAL3, ISO/IEC 29115 level 3 or 4. The protocol does not accept anonymous signers. We do not maintain our own classification, we rely on existing sovereign standards. Identity stays private; only a hash is published on the registry.

2. ENTITY 02 · WHAT

   Digital content.

   Cryptographic fingerprints computed client-side: exact SHA-256 plus perceptual hashes robust to recompression, trimming, and re-encoding. No content is stored on the registry. The content stays with whoever publishes it.

3. SIGNATURE · THE OBJECT OF THE PROTOCOL

   Cryptographic, by graduated role.

   The signature links identity to content and specifies the role of the signer: author, actor, witness, broadcaster, or reclaim. This is what the protocol defines. Everything else (identity on one side, content on the other, registry as archive) is infrastructure.

4. ARCHIVE LAYER · WHERE

   Public blockchain.

   The signature is recorded on a public blockchain to guarantee immutability and censorship-resistance. The protocol is chain-agnostic; multiple profiles are possible (Polygon, Base, Ethereum, Solana). No actor can erase what has been signed.

On top of this structure, one conceptual move: the graduated signature by role includes Reclaim, which lets a verified identity sign an existing piece of content to refute it, with context and explanation. The rebuttal automatically propagates to every derivative of the falsified content and catches the fake wherever it circulates.

The open-source commitment

A standard cannot be proprietary. A trust protocol cannot be opaque. A tool against disinformation cannot sit under the control of a single actor.

The Alethea Protocol specifications are published under Creative Commons Attribution 4.0. Anyone can cite, implement, or extend the spec, provided they attribute authorship.

The reference implementations are published under AGPLv3. Any publicly distributed modification must stay open. This license blocks closed commercial capture of the protocol.

The protocol is chain-agnostic, identity-provider-agnostic, and distributable by anyone. Just as HTTPS does not depend on a single operating system, Alethea does not depend on a single infrastructure. The spec defines the structure of a signature and the shape of a registry, not where these objects physically live.

Alethea does not accept anonymous signers. Every signature engages a human identity verified by an internationally recognized trust standard: eIDAS, NIST 800-63, or ISO/IEC 29115. We do not create our own classification; we reference existing sovereign standards. The protocol is prescriptive on identity quality without becoming a classification authority itself.

Long-term governance will be transferred to a neutral foundation once the number of contributors and adopting organizations justifies it. Linux Foundation, Mozilla, Wikimedia model. Any major spec evolution goes through a public RFC process.

The call

This protocol has value only if adopted. We call on:

• ·developers to implement, stress-test, critique the spec, ship SDKs on their platforms;
• ·researchers in cryptography, provenance, and disinformation to challenge the technical and conceptual choices;
• ·media and newsrooms to integrate Alethea verification into their editorial workflow;
• ·public institutions, elected officials, and communications teams of local governments to become pilot signers;
• ·European and national legislators to recognize signing public content as a democratic requirement.

The code is open. The spec is open. The governance will be open. The protocol is yours as much as ours.