Your control fired.
Did it work?
Detection is measured in seconds. Proof is signed in minutes. We send a live adversary agent at each control and sign only the result that matches what was committed before the run.
control fired / logged
expected = actual
Every attack, read out.
Firing is not working.
Proof of Control
The control fired. A SIEM entry, an alert, or an audit record says so. Nothing shows whether the agent got through.
Proof of Efficacy
The control worked against a real adversary technique, in a customer-like environment, with the expected result committed before the run.
Chaotic.core™ produces both, and the signed record only counts when efficacy matches what was committed.
Commit. Attack. Measure. Sign.
-
01
Commit
Expected outcomes are locked before anything runs. No moving the goalposts after the agent has fired.
-
02
Attack
Adversary agents run real techniques with no industry mappings against the controls in an isolated environment.
-
03
Measure
Actual outcomes are captured as they happen: did the control stop the agent, or did the attack get through?
-
04
Sign or withhold
Expected matches actual, so the record is signed. Divergence is withheld until the control is fixed and retested.
In the lab. Or live, alongside real traffic.
Arena
Model the target locally, then spin up a full arena in the cloud. Good for pre-deploy checks and vendor evaluations.
Continuous
A witness runs in production with no start and no finish. Proof updates as the control changes and as new agents appear.
Open specs.
Controlled code.
The Proof Protocol specs are open on GitHub, so anyone can check a signature, read the declared AI contribution, and see who funded the work. The code that produces the proofs stays controlled.
Read the specsStop attesting.
Start proving.
Vendors proving a product. Buyers verifying one. Insurers pricing the gap. Send us one control to start.