Briefing · October 2026

Prove the real, at the sensor

Science journals and insurers both ask the same question about a picture: is it still what the sensor saw? A fragile seal answers it, and ends the cat-and-mouse game between fakes and detectors.

15 minutes · subtitles in the video30-second trailer
Start a pilot conversationFORTRESS: science integrity

Versions for three houses

The same story, with one slide written for the reader. These are briefings, not partnerships: none of these houses has agreed to anything.

For Nature

Today two AI tools pre-screen submissions, and both only flag; a person decides.

Open the briefing (no partnership) →

For Carl Zeiss

Sealed at acquisition in ZEN, carried inside the CZI file, checked by an open verifier.

Open the briefing (no partnership) →

For Allianz

In June 2026 your UK team found an AI-added bumper crack on a claim photo; the accident never happened.

Open the briefing (no partnership) →

Check it yourself

Do not take our word for it. The open verifier and the evidence page are public: verifier · evidence · Proof of Agent Work.

Honest status

What is real, and what is not.
  • Working today: the fragile seal as software and browser SDK. Lab result: 0 false alarms on benign changes, 6 of 6 kinds of tampering caught (TRL 6).
  • In development: native device attestation on Android and iOS.
  • A proposal, built with a partner: the seal inside secure chip, image-pipeline firmware and sensor readout.
  • Not claimed: the seal proves that a picture was captured live and has not been changed since. It does not prove that a scene is true, and it does not detect AI by guessing. Patents: US provisional applications, filed, not yet examined.