For insurers and reinsurers

AI agents settle claims. Deepfakes file them. Check both yourself.

destill.ai gives you two kinds of evidence: a signed receipt for every run of an AI agent, and claim photos sealed at the moment of capture. Your own people check both in seconds, without having to trust us.

The site in about a minute. Voice: neural text-to-speech. The check in the film is the real one below.
Two doors for insurers

Which problem is on your desk?

Live check

Try to cheat.

This is a really signed receipt, verified in your browser right now; the claim in it is example data. Raise the payout, erase the human approval, hide the fraud score, and watch the check fail.

More doors

The same evidence, for your role.

All markets

Ad fraud and AI shopping agents

Which copy of your social creative leaked, and which billed event really came from your agent.

Deepfakes and proof of real

Photos sealed at capture: insurance claims, research images, identity checks, newsrooms.

Anti-piracy and leak tracing

One invisible key per recipient: which screener, licensee or press-kit copy leaked.

EU AI Act: evidence for AI agents

A signed receipt for every run of your AI agent, and a screen for prompt injection in front of your model.

How it works

Three modules, and what a check does not prove.

POAW

Proof of Agent Work signs each step an AI agent takes into a receipt anyone can verify offline.

How a receipt works →

Fortress

Fortress hides an invisible, key-bound mark in images and media and reads it back to show which copy leaked.

Marks and seals →

AEGIS

AEGIS screens prompts and tool inputs for injection attempts on a plain CPU before they reach your model.

Known gaps: role-play and story-style jailbreaks, false alarms on non-English text, paraphrased attacks. We publish no detection rate until a reproducible public rerun exists ().

What a check proves — and what it does not