WEBVTT

NOTE Synthetic voice (text-to-speech). Text: films/data/films.json, timing: films/data/manifest.json (tools/film-captions.mjs).

00:00:00.000 --> 00:00:02.600
[Synthetic voice (text-to-speech)]

00:00:03.100 --> 00:00:04.483
Friday afternoon.

00:00:04.483 --> 00:00:09.038
You're on the fraud desk, and three claims look perfect.

00:00:09.038 --> 00:00:10.014
Too perfect.

00:00:10.014 --> 00:00:13.756
The photos are sharp, the damage is plausible.

00:00:13.756 --> 00:00:15.220
Is any of it real?

00:00:16.620 --> 00:00:23.868
A photo can be generated, an old photo can be reused, or a screen can simply be photographed again.

00:00:25.268 --> 00:00:29.759
Fortress seals the photo inside your app, at the moment of capture.

00:00:29.759 --> 00:00:38.204
In our tests, screenshots and AI redraws broke the fragile seal every time, so the claim is flagged before the money goes out.

00:00:39.604 --> 00:00:43.581
Untouched originals read clean, with no false alarms.

00:00:43.581 --> 00:00:44.932
Edits are flagged.

00:00:44.932 --> 00:00:50.860
And it runs on the device, without sending your customers' photos to any cloud.

00:00:52.260 --> 00:00:55.023
It works for photos captured in your app.

00:00:55.023 --> 00:01:01.829
It cannot vouch for a photo taken somewhere else, and a light JPEG re-save is deliberately tolerated.

00:01:01.829 --> 00:01:05.940
That's why we pair it with a signed hash for bit-exact proof.

00:01:07.340 --> 00:01:11.191
Pay real claims faster, and flag the fake ones.

00:01:11.191 --> 00:01:14.222
License Fortress for your claims app.

00:01:14.222 --> 00:01:16.844
Destill dot A I, slash fortress.

