M

Case Study

Ensuring High-Quality Healthcare Through Penetration Testing

In-house AI testing engine

Katana: AI penetration testing, led by human experts

The speed of AI. The instinct of expert penetration testers.

Katana is the AI testing engine we built in-house to think like our best penetration testers, tuned on years of Samurai testing experience. Our consultants use it to cover more ground, faster, then apply the creativity and judgement only experienced pentesters bring. It isn't an off-the-shelf scanner or a bolt-on product. It's how we test.

Prefer no AI? That's absolutely fine.

If you'd rather not use AI, we won't. Our consultants will test entirely by hand, with the same team, the same methodology and the same quality of report. It's your choice, agreed at the scoping call.

Built from the ground up

Not a chatbot with a security wrapper

Built with real pentester experience and research-led academic rigour.

Plenty of "AI pentesting" tools are a general-purpose language model with a thin security layer on top. Katana isn't. We built it from the ground up to think the way our top testers think: how they plan an attack, where they look first, and how they follow a weak signal until it becomes a real finding. It's tuned on years of Samurai testing experience and our consultants' methods, so it reflects how our team actually tests, not what a generic model guesses a test should look like.

Katana was designed by a team with 25 published CVEs and is led by Dr David Day, one of the first people in the UK to earn a hard technical PhD, in detecting existing and polymorphic shellcode on the wire, and a senior lecturer and researcher in cyber security for 13 years. David is a Special Officer with the NCA's National Cyber Crime Unit and a Chartered Cyber Security Professional with the UK Cyber Security Council.

  • 25 published CVEsDesigned by the same team, led by Dr David Day
  • Human-verifiedNo AI finding reaches your report until a tester has confirmed it
  • Your choiceUsed only with your permission, or not at all

How it works

How Katana works

Katana runs like a pentest team made of AI agents:

  1. 01Supervising agentsAct as team leaders. They plan the test, assign the work and keep it on scope.
  2. 02Hunting agentsAct as pentesters. They probe your systems for weaknesses and try to exploit them.
  3. 03Proofing agentsCheck every finding and remove false positives, before a human tester verifies each one.

Permission first

Safe, controlled and only with your permission

We only use Katana on an engagement when the client has agreed to it. You choose human-led, hybrid or fully Katana testing at the scoping call, and if you'd rather we didn't use AI, our consultants test entirely by hand.

When Katana is used, it works within the same agreed scope, rules of engagement and testing windows as our human testers. Its supervising agents keep every action on scope, and our CREST-accredited consultants oversee the test from start to finish. Nothing is tested without written authorisation, and you always know who to call.

Every finding Katana raises is verified by one of our human testers before it reaches your report, and our consultants often explore it further to show how far an attacker could really take it.

Your data is never used to train Katana.

Katana

What Katana brings

  • Tested the way AI attackers work. Malicious hackers are already using AI agents to find and chain weaknesses at speed. Katana uses the same techniques, under strict client permission and with every finding verified by our testers, so it's the clearest way to find out whether you're susceptible to AI hacking agents before someone else does.
  • Speed and breadth. It works through large attack surfaces fast, checking every host, endpoint and input.
  • Thinks like our testers. It's built around our consultants' methods and tuned on a decade of our own testing experience, not generic training data.
  • Human-verified, every time. No AI finding reaches your report until one of our testers has confirmed it.

Human testers

What our human testers bring

AI covers the ground, but people still find the issues that matter most.

  • Creativity. They find weaknesses no scanner or model is looking for.
  • Chaining. They link small issues into a real path to your data.
  • A red-team mindset. They think like a determined attacker with a goal, not a checklist.
  • Findings that develop over time. They catch issues that only show up across sessions, states or time.
  • Business context. They judge what actually matters to your organisation.

Your choice

Choose your approach

  • Hybrid Most popular

    Katana covers the ground, and our consultants test deeper and review every finding (with your agreement).

  • Fully human

    Our consultants do all the testing and write the report.

  • Fully Katana

    The fastest option. Katana does the testing, and our human testers verify every finding before the report is issued (with your agreement).

Whichever you choose, it's the same Samurai team, the same methodology and the same accountability.

The process

How an engagement runs

  1. Scoping call

    You tell us what needs testing, and we agree the scope and the right mix of human, Katana or hybrid testing.

  2. Clear quote

    You see the consultant days your scope needs and the day rate behind them.

  3. Testing

    Our consultants, with Katana if you choose it, test and chain findings the way a real attacker would.

  4. Report

    We aim to deliver it within 5 working days of testing finishing, with severity and remediation advice for each finding.

  5. Retest

    Once you've fixed the issues, we retest, priced by the number of findings to re-verify.

Trust

Why Samurai

  • NCSC Assured Cyber Security Consultancy
  • CREST accredited penetration testing
  • ISO 9001 certified (UKAS)
  • Cyber Essentials Plus certified

“We were astonished by what Samurai's Black Dragon assessment revealed, even though we thought we had everything covered.”

Rezolve AI · Case study

“Samurai takes the time to understand our needs and helps us to be an enabler for Flagship Group's mission.”

Flagship Group · Case study

Questions

FAQs

Is AI penetration testing as good as a human tester?

On its own, no. That's why we combine the two. Katana is excellent at speed and coverage, but chaining issues together, creative attacks and business judgement still need experienced people. Our hybrid approach gives you both.

Is Katana just ChatGPT or a scanner?

No to both. Katana isn't a general-purpose chatbot with a security wrapper, and it isn't a signature-matching scanner. It was built to reason like our top testers and is tuned on years of our own testing experience. And unlike a scanner, nothing goes into your report unverified. Every finding is confirmed by a human tester, and often explored in more detail.

Will you use Katana without asking me?

No. We only use Katana with your permission, agreed at scoping and written into the scope of work. You can choose a fully human test instead.

What if I don't want AI anywhere near my systems?

Then we won't use it. Fully human testing is always available, and you lose nothing on quality or accreditation.

Is Katana safe to run against my systems?

Yes. Katana stays inside the scope, targets and testing windows you approve, and our consultants supervise it throughout, just as they would a human-led test. We'll agree any sensitive systems or exclusions with you before testing starts.

Is my data used to train Katana?

No. Your data is never used to train Katana.

Will a Katana-assisted test meet compliance requirements?

Every engagement is delivered by our CREST-accredited team and produces a full report with severity ratings and remediation advice, suitable as evidence for frameworks such as ISO 27001 and Cyber Essentials.

How much does it cost?

It depends on scope. Our penetration testing cost calculator shows our day rate and typical consultant days, or you can book a scoping call for a fixed quote.