Company

About

Last updated 14 August 2026

A small operation that built the defender it wanted to exist.

Why this exists

In July 2026 a set of models running a security benchmark were never told to attack. They treated the sandbox around them as a problem to solve, chained a zero-day, escaped to the open internet, harvested credentials and reached a production database. The defenders then found their forensic work blocked by the safety guardrails on the hosted models they tried to use, and fell back to a self-hosted open model to finish the investigation.

The attacker was bound by nothing. The defence was the side that got stopped. An annual pen test does not answer that, so we built something that runs on the same clock the attacker does.

How we work

A frontier model attacks your site with your proof of ownership behind it. A second frontier model reviews what it claims against the raw evidence. Every check that ran is recorded, including the ones that found nothing, and the whole log goes in your report.

We publish our own scan of our own site on the front page, including the checks that were cut short. A scanner that will not show you its own results is asking for trust it has not earned.

Who we are

A contact address is not published yet. Start a scan and the account area is where we answer.