Proving What’s Real at the Blockathon

A convincing fake of a face, a voice or an official message can be produced in minutes. Proving that the original is genuine is still slow, manual and easy to dispute. As AI-generated content spreads, that imbalance becomes a problem for anyone whose work depends on being believed: creators, aid organisations and emergency responders among them.

Three teams took on that problem at the FirstBlock-athon in Stockholm. Each worked on a challenge brought by an organisation that lives with it every day. They reached the same answer by different routes, and that answer works as a blueprint for any organisation that needs to prove what is real.

Three organisations, three challenges

The challenges came from AIMS-Management, the Swedish Red Cross and UNICEF. AIMS-Management asked how creators can protect their identity when their face and voice are easy to copy. The Red Cross asked how people in a crisis can know a message truly comes from them. UNICEF asked how to get credible proof from the field that an aid programme was carried out.

Neuro was the technical provider for the blockathon and made its technology available for the builders to work with: infrastructure for digital identity, digital signatures and verifiable records.

One term comes up in all three projects: the digital fingerprint. A digital fingerprint (a hash) is a short code calculated from a file. The same file always produces the same code, and changing a single pixel or word produces a completely different one. Storing the fingerprint makes it possible to prove later that a file is exactly the one that was approved, without storing the file itself.

OWND: proof of what a creator agreed to

Creators earn a living from their face, voice and style. When AI can copy all three, a brand or a viewer has no easy way to tell which content the creator actually approved. The usual response is detection: build tools that spot deepfakes. Detection is an arms race, and every improvement in generation makes it harder.

Team OWND chose to document consent at the source. When a creator approves a piece of content, the file gets a digital fingerprint and the approval is tied to that exact file. If anyone edits the file afterwards, the fingerprint no longer matches and the approval no longer applies. No personal data needs to be stored on the blockchain to make this work.

Their demo walked through the full flow a creator would go through: an identity check, a live photo, a voice sample and an approval step that defines which rights a brand or UGC (user-generated content) company gets over the creator's likeness. The result is a record that answers the questions a dispute would raise: who agreed, to what, and for which file.

FirstMark: crisis messages people can verify

In a crisis, information moves fast and false information moves faster. A fake message that looks like it comes from the Red Cross can send people to the wrong place or make them ignore the right advice. AI-generated text and images make those fakes cheaper to produce and harder to spot.

Team FirstMark built a system for verifiable crisis messages. The sender signs each message digitally and sets how long it should be valid. The recipient can open a verification receipt showing who sent the message, when it was published, when it expires and the digital fingerprint of the message itself. The fingerprint is registered on a blockchain, while the message content can be stored elsewhere.

The expiry date is the detail worth noticing. Crisis information goes out of date quickly, and an old message that was genuine can mislead as much as a fake one. Building validity into the message means a recipient can check both where it came from and whether it still applies.

U-Proof: evidence that aid arrived

Aid organisations answer to donors, governments and the people they serve. Each of them wants to know the help actually reached its destination. Reports from the field often travel through several hands before they arrive, and every step adds room for error or doubt.

Team U-Proof built a prototype that moves the evidence closer to the source. It is a mobile web app where the person receiving help takes a photo inside the app when food or other support is delivered. The time and approximate location are added automatically. UNICEF can then review and approve each submission.

The photo stays in UNICEF's own systems. Only its digital fingerprint, together with the location area and time, is registered on a blockchain. That keeps sensitive images under the organisation's control while creating a record that cannot be quietly changed later. The team also described U-Proof as a possible mobile-first extension to UNICEF's U-Report platform.

The pattern behind all three

A creator's consent, a crisis message and a photo from an aid delivery have little in common on the surface. The three teams still built the same four-part structure, and it applies far beyond these cases.

The first part is fingerprinting the thing itself. Every project starts by calculating a digital fingerprint of the exact file or message. That fingerprint is what makes any later change visible.

The second part is binding it to a verified identity. A fingerprint proves the file is unchanged. It takes a verified identity and a digital signature to prove who approved it, who sent it or who received it.

The third part is keeping personal data where it belongs. None of the teams put photos, messages or personal details on a blockchain. The data stays with its owner, and only the proof goes on the ledger. That choice matters under data protection rules such as GDPR, where personal data written to a permanent ledger would be hard to correct or delete.

The fourth part is defining the terms. OWND records which rights a brand gets. FirstMark records how long a message is valid. U-Proof records that UNICEF has reviewed and approved a submission. The proof means more when it states what was agreed, for how long and by whom.

Any organisation can test its own processes against these four parts with one question: if someone challenged this document, image or message, could you show who approved it, which exact version, and when? If the honest answer depends on email threads and memory, the process is a candidate for the same structure.

From prototype to production

The three projects are prototypes, built within the format of a blockathon. The principle they share already runs in production.

W Social, a European social platform, uses Neuro's identity layer to verify that every user is a real person, without saving their identity data. It is the same structure the teams arrived at: a verified identity, proof that can be checked, and personal data kept out of places it does not belong.

Read the W Social customer case to see how it works in practice.