Capability Systems

Agentic AI Security Testing

Could an attacker reach your equipment from the internet?

In autumn 2026, attackers took over refrigeration in supermarkets across Europe, in some cases turning freezers into heaters overnight. France’s cyber authority warned that the controllers targeted could be reached from the internet, and many were older models that no longer received security updates. Food manufacturers run the same kinds of controllers. Skwiz, our agentic AI security tester, searches for exactly this kind of weakness. Its team of specialist AI agents maps everything you expose to the internet and hunts through it the way an attacker would, deciding for itself which test to run next. Controllers like the ones in those attacks, reachable from the internet, are exactly what it looks for.

We built Skwiz to test our own software and services first. Now we use it to check yours.

What Skwiz Checks

Only the systems you authorise, searched from the outside for the weaknesses an attacker would look for.

Equipment and control panels on the internet

We check the internet addresses you name for management and control interfaces that should never be public, such as the login page of a refrigeration, HVAC or building controller, or a device's admin console. These are exactly what attackers search for.

Your websites and portals

Customer, supplier and staff portals are checked for known weaknesses, exposed files and admin pages, and weak security settings such as missing HTTPS protections, using established scanners.

Your code and dependencies

If you share your source code with us, we look for passwords and keys committed to it, out-of-date components with published vulnerabilities, and common coding mistakes.

Safe by Design

Skwiz tests carefully, within limits you agree in advance.

You authorise every test in writing

Skwiz will not run without a signed scope that names who authorised the work and exactly what it may test. Anything outside that scope is refused, and blocked at the network as well, so a mistake cannot reach someone else's systems.

It looks, it does not attack

By default Skwiz observes. It does not try passwords or run exploits. Any more active test needs your explicit approval first, and testing a live production system also needs a named approver and an agreed time window.

Every action is recorded

Each step Skwiz takes is written to a tamper-evident audit log, so you can see exactly what was checked and when.

Agentic AI driving proven tools

Skwiz's AI agents plan the testing, decide what to look at next and run established open-source security tools, including nmap, Nuclei, OWASP ZAP, Semgrep, Trivy and gitleaks. Before it reports a weakness as confirmed, Skwiz looks for a second, independent sign of it. It never writes attacks of its own.

Your Free Exposure Check

We check your internet-facing systems at no cost. Code reviews and wider testing are quoted separately if you want them.

  1. 1

    Tell us what to check

    The domains and internet addresses your business uses: your website, portals, remote access and any equipment connected to the internet.

  2. 2

    Authorise it

    You sign a short scope document naming exactly what we may test. Nothing outside it is touched.

  3. 3

    We run the check

    Non-destructive, at a low rate and at a time you agree. A device that cannot cope with a careful check from outside would not survive the scanning attackers carry out every day, and if we see one we will tell you straight away.

  4. 4

    We talk you through the results

    You receive a report listing anything exposed, how serious it is and the evidence for it, which your IT team or supplier can act on. We go through it with you.

What It Does Not Cover

Skwiz tests what can be reached from the internet, and the code you share with us. It is not an assessment of the control network inside your plant, and it does not test the equipment itself. If we find something that needs specialist operational-technology work, we will tell you and help you find the right people.

Find out what an attacker can already see

Tell us which systems to check. We will agree the scope with you in writing, run the check at a time that suits you, and talk you through what we find.