THE MISSING LAYER
Generation became instant. Rights did not.
A face, a voice or a character can now travel through dozens of models, studios, agents and distribution platforms in minutes. The underlying permission still lives in contracts, email threads, PDF attachments and private databases. This gap is where misuse, ambiguity and avoidable risk begin.
Watermarks and provenance help us understand how media was produced or altered. They do not, by themselves, prove that the person, rightsholder or authorised representative approved this specific use. Transparency answers “what is this?” Rights infrastructure must also answer “may this happen?”
The future of synthetic media cannot rely on trust by implication.
A REGULATORY INFLECTION POINT
2 August 2026 is a starting line, not a finish line.
From 2 August 2026, Article 50 of the EU AI Act applies transparency obligations to providers and deployers of certain AI systems and AI-generated or manipulated content. Machine-readable marking and disclosure are becoming part of the operating environment for generative media.
This is an essential shift. Yet disclosure alone does not establish consent, mandate, contractual scope, expiry, revocation or permission for a particular territory and channel. Compliance can make synthetic content visible. Trust requires the authority behind it to become inspectable too.
We should not wait for a crisis to make rights machine-readable.
THE PROOF MODEL
Four decisions. One chain of authority.
Proof separates durable rights from temporary execution and later distribution. A signed credential is not a universal pass. Every action remains contextual, limited and accountable.
Convert verified authority, explicit consent and agreed terms into a signed, portable and revocable licence credential.
Evaluate the requester, reference asset, provider, model, script, purpose, territory, channel, time and usage limits before generation.
Register the final output, cryptographic digest, provider evidence and compatible provenance reference.
Return an explainable allow, deny or review decision using signature, status, policy, asset binding and distribution context.
DESIGNED TO CONNECT
Complement standards. Do not create another silo.
Proof is designed to work with the ecosystem already forming around synthetic media. C2PA and Content Credentials can carry provenance and tamper-evident assertions. Proof adds the operational rights chain: who had authority, what was approved, whether it remains active and whether the present use is within scope.
The core protocol works without a blockchain. Where organisations need an additional timestamping or external audit layer, anchoring can be optional and chain-agnostic. No vendor, ledger or generation model should become a permanent point of dependency.
One proof should travel across every workflow, not trap rights inside one platform.
OUR PRINCIPLES
People remain the source of authority.
A platform may request a licence. It must never invent consent or self-issue another person’s rights.
Purpose, identity, script, provider, model, territory, channel, duration and volume must be expressible and enforceable.
Holding a credential is not enough. Each generation and distribution action must be evaluated in context.
Expiry, suspension and withdrawal are live protocol states, not footnotes in an old contract.
Allow, deny and review outcomes must produce stable reasons and auditable receipts.
Verification should reveal the minimum necessary evidence without exposing private contracts or identity documents.
Rights must survive a change of studio, model, provider, publisher or technical stack.
External anchoring may reinforce trust, but the protocol must remain useful, neutral and operational without it.
AN OPEN INVITATION
A protocol earns trust through scrutiny.
Proof is currently an early preview and interactive whitepaper. It is not yet a production issuer, an authorisation network or an industry standard. The identities and activity shown in the product are fictional. The architecture, questions and ambition are real.
We are opening this work now because rights infrastructure cannot be designed by one company alone. It requires talent and rightsholders, agencies, brands, studios, AI providers, publishers, standards experts, policymakers, security researchers and civil-society voices around the same table.
We are looking for rigorous review: what is missing, what should remain out of scope, what must be interoperable, and what governance would make this worthy of adoption.
The goal is not to own trust. It is to make trust portable.