AI security architecture for regulated organisations
Approve AI on evidence, not exceptions.
Independent security architecture for AI, agentic systems and emerging technology. I threat-model what is actually being built, design proportionate controls, and produce the evidence your risk, audit and approval forums will ask for.
Banks, asset managers, custodians, the fintechs that sell to them and the law firms that advise them.
CISSP · CCSP · 10 years in cybersecurity · Edinburgh and UK-wide · Delivered personally
When to bring me in
The business has bought an AI tool and needs it approved.
I prepare the case your security and risk reviewers need: threat model, controls and evidence. The decision stays with them.
You sell AI into banks, and deals stall at their review.
I have run those reviews from the buyer side. I build the answer sheet for the questions a bank's reviewers will ask, and work through it with you before they ask.
A design or platform needs an independent review before go-live.
A threat-model-led review ending in an approve, approve-with-conditions or reject recommendation.
Architecture review and assurance →Agent tool-access review →
A programme needs security architecture and governance in the room.
Senior design authority while decisions are still being made, from AI standards to cloud target states.
Running a fintech or smaller firm without a security function? Start with an AI security baseline →
A consultancy or partner that needs AI security depth on an engagement? Discuss a partnership →
Starting points
Fixed-scope engagements for the questions that come up most. Each is a fixed fee, agreed in writing after a short scoping call.
AI tool approval pack
Typically 2 to 3 weeks
What would it take for this tool to be approved, and what does the evidence look like?
AI vendor assurance pack
Typically 2 to 3 weeks
What will a bank's review ask of our product, and how strong are our answers?
Microsoft 365 Copilot readiness review
Typically 1 to 2 weeks
What will people be able to find on day one that they shouldn't, and what has to be contained first?
Agent tool-access review
Typically 1 to 2 weeks
What can this agent actually reach and change, whose authority does it act on, and what would stop it?
Every engagement runs the same way: agree the question and the decision criteria, gather the evidence, analyse, and hand over. How an engagement runs →
Track record
Work I led before founding Cairn, described by sector and scale. Clients and former employers are not named, and no organisation's security posture is described.
Establishing AI security governance at a global asset manager
Asset management · Institutional scale, UK-regulated
An enforceable AI security standard, with gateway controls behind it, in place before the main rollout.
Blockchain security standard for an institutional custodian
Custody and institutional banking · Global custody and asset servicing
A blockchain security standard giving the firm a defined basis for adopting digital assets.
See what you would receive
An illustrative executive decision brief from an AI tool approval pack, for a fictional organisation: the risks, the controls required before go-live, the conditions, and the recommendation.
Services
AI & agentic AI security architecture
Secure AI systems before they reach production, with practical threat modelling, gateway controls and governance designed for regulated environments.
AI security standard · Threat model · Gateway control design · Agent tool-access review · Approval pack for a purchased tool
Cloud & hybrid security architecture
Security designed into cloud migrations, platform builds and post-merger integrations early enough to shape them, and evidenced well enough to pass audit.
Security standards & governance
Standards, control frameworks and governance processes that survive contact with an audit and with delivery teams.
Architecture review & assurance
Independent security review of designs, platforms and third-party systems, ending in a recommendation you can defend.
Independent advice. Senior expertise. No account team.
Cairn Cyber Advisory is an independent AI security practice in Edinburgh, run by me, Gary Johnston: a security architect with ten years in cybersecurity across institutional banking, asset management, legal services and telecommunications.
The work is led and delivered by me, so you get senior advice directly rather than through an account team. I sell no products and take no vendor referral fees.
Recent writing
The problems that come up repeatedly, written up in enough detail to be useful to someone solving them without me.
Before you switch on Copilot: the permissions problem it inherits
· 5 min read
Copilot creates no new access. It makes the access people already have usable, and years of oversharing become one question away. Readiness means containing the worst of it, not cleaning the whole tenant first.
Your model provider is an ICT third party. Most registers don't say so.
· 6 min read
An LLM behind a business process is an ICT third-party dependency under DORA and the UK outsourcing rules. It usually arrives without a procurement event, so it is usually missing from the register, the contract review and the incident process.
ISO 42001 readiness without building a second bureaucracy
· 5 min read
ISO/IEC 42001 asks for a management system for AI. With ISO 27001 in place most of the structure exists already; readiness is about the parts that are new.
Start a conversation
Send me an outline of the problem and your timescale. I normally reply within one working day and will tell you directly whether I think I can help.