CREST Certified · AI-Augmented · Anthropic Approved

Custodian Consulting

Offensive Security. Defensive Confidence.

We've done the impossible with AI — a fully autonomous offensive engine that compromises most machines in under 10 minutes, and we're Anthropic-approved to do it.

custodian@c2:~$ ./custodian-ai --target client-scope.yml --engine kill-chain --autonomous
[*] Custodian AI Engine — recruiting attack assets | Anthropic-whitelisted
[+] Fleet online — 2x Linux Operators + GPU cracking rig
[*] Autonomous kill-chain engaged — enumerate → research → exploit → verify
[+] Live exploit sourced — Apache 2.4.49 path traversal → RCE
[!] Foothold landed — www-data @ target-01 | priv-esc via SUID → root
[!] Hashes offloaded — cracked on GPU in 41s
[+] Host fully compromised — root in 8m 12s (autonomous, no operator)
[*] Self-learning loop — feeding new attack surface back to recon
[+] Full domain compromise achieved in 4h 23m
custodian@c2:~$ generate-report --classification CONFIDENTIAL --output engagement_2026-Q1.pdf
[+] Executive report generated — 47 findings | 6 critical | 12 high
LIVE ENGAGEMENT The Custodian AI engine running a live autonomous kill-chain across multiple targets — recon, live exploit research, exploitation, independent verification and a self-looping failure-analysis stage.
The Custodian AI engine driving a live, autonomous, multi-target kill-chain

We break things before the adversaries do.

Custodian Consulting is a UK-based offensive security firm built by penetration testers, for organisations that refuse to gamble on their security posture.

Our team holds CREST, OSCP, and OSCE certifications. We have delivered engagements across NHS trusts, central government departments, financial services, and critical national infrastructure. Every finding we report is exploitable, every recommendation is actionable.

We do not sell fear. We sell evidence. Our reports give your board the clarity to make informed decisions about risk, and your engineers the technical detail to fix what matters.

CREST
OSCP / OSCE
Cyber Essentials Plus
500+
Engagements Delivered
99.8%
Client Retention Rate
15+
Years Combined Experience
0
Breaches Post-Engagement

Autonomous offensive AI that compromises most machines in under 10 minutes.

We have weaponised AI for offensive security. Custodian's engine runs a fully autonomous attack pipeline — recruiting its own infrastructure, driving a relentless kill-chain, and learning from every engagement. What once took a consultant hours, it does in minutes, then hands you a board-ready report of exactly how it got in.

< 10 min
from first packet to full machine compromise — autonomously

Recruiting Attack Assets

The engine builds its own arsenal on demand. It provisions, clones and orchestrates a fleet of attack assets — distributed Linux operators and dedicated GPU cracking rigs — recruiting exactly the firepower each target needs and retiring it when the job is done.

Autonomous Kill-Chain

A persistent, looping kill-chain — enumerate, research live exploits, weaponise, exploit, verify — that drives toward the objective without a human in the loop. When an attack path stalls it reasons about why, pivots, and tries harder until the target falls.

Self-Learning

Every command, success and dead-end is logged and fed back into the system. The engine distils each engagement into reusable methodology, so it gets faster, sharper and harder to stop with every machine it touches — your test benefits from every test before it.

Anthropic Whitelisted

Our engine is built on Anthropic's Claude and operates under approved, whitelisted access for offensive security work — sanctioned AI, applied to authorised engagements only. Real adversaries already use AI; we make sure you are tested by it first, responsibly and under contract.

What We Deliver

End-to-end offensive security services, from scoping through to remediation validation. Every engagement is bespoke, every report is board-ready.

Built for organisations that take security seriously.

We are not a reseller with a scanner. We are hands-on-keyboard operators who understand your threat landscape.

CREST Accredited

Certified to test UK government and CNI systems. Our testers hold individual CREST qualifications at CRT and CCT level.

Board-Ready Reporting

Executive summaries your leadership team can act on, technical appendices your engineers can build from. No filler, no recycled scanner output.

Manual-First Methodology

Automated tools find the low-hanging fruit. Our testers chain vulnerabilities, abuse business logic, and identify the attack paths that matter.

Proven Track Record

Trusted by NHS trusts, local authorities, MOD suppliers, and FTSE-listed organisations. We operate under NDA with the discretion your sector demands.

Rapid Turnaround

Most engagements scoped within 48 hours and scheduled within two weeks. Emergency incident response available with same-day mobilisation.

Remediation Validation

We do not just find problems and walk away. Every engagement includes a free retest window so you can verify your fixes hold under pressure.

Ready to test your defences?

Tell us what you need secured. We will scope the engagement, provide a fixed-price quote, and schedule your test — typically within two weeks.

Start a Conversation