Security & Governance

AI risks, governance and controls: Attacks: Prompt injection, Jailbreaks, Data leakage; Governance: AI policy, ISO/IEC 42001, EU AI Act; Controls: Least privilege, Allow-lists, Audit trailAI riskAttacksPrompt injectionJailbreaksData leakageGovernanceAI policyISO/IEC 42001EU AI ActControlsLeast privilegeAllow-listsAudit trail

Bringing AI into your business brings great benefits, and risk too. Securing it takes a blend of governance, policy and technical understanding, and a strategy for building AI into your systems and processes. We can provide Chief AI Officer leadership.

Services

AI security reviews
An independent review of your LLM features, agents and AI suppliers: prompt injection, data leakage, over-privileged agents and model supply chain risk. You get fixes ranked by impact, before your users find the gaps.
Security architecture
Threat models, trust boundaries, identity and data flows designed in from the start, so security is part of how the system works rather than a report at the end.
Cloud architecture reviews
An independent review of how your cloud estate is designed: identity and access, network boundaries, where data is stored, logging, backup and recovery. You get the risks that matter most, and a plan to fix them in order.
Cloud penetration testing
Authorised, hands-on testing of your cloud environment, approached the way an attacker would: exposed services, misconfigured access, weak secrets and routes from one account or service into another. Every finding comes with evidence and a fix.
Cloud security posture management (CSPM)
Continuous checks of your cloud configuration against good practice and recognised standards, so new misconfigurations are caught as they appear, not at the next audit. We set it up, tune out the noise and help you act on what matters.
Secure code review
Close review of the code that matters most, including code written by AI assistants. Findings are ranked by real-world risk and come with clear fixes your team can ship.
Chief AI Officer support
Senior AI leadership, part time, for organisations that need it without a full-time hire. We set the AI strategy, decide where AI is worth using, keep the risks in view and report to your board.
AI policy and ways of working
Acceptable use, approved tools, rules for client and personal data, and review practices for AI-written work. Written for how your teams actually work, so people follow them. Governance for Claude and Claude Code.
AI training
Practical sessions for staff, engineers and leadership: using AI tools well and safely, building with AI, and spotting where it goes wrong.
ISO/IEC 42001 readiness
Getting your AI management system ready for certification. Launching soon.
Other security services
Second opinions, supplier and third-party risk reviews, and security advice for leadership teams. If it isn't listed, ask.

Your staff already use AI. Bring it under control.

Most businesses didn't decide to adopt AI. Their staff did. We find out what's in use, agree with you what's allowed, and set up Claude and Claude Code so the controls hold.

AI usage governance

For any business

We find out which AI tools your staff already use, including the ones nobody approved. We write a data-handling policy people can follow, then set up Claude Team or Enterprise with the right access and admin settings from the start.

You get:

  • A list of the AI tools in use
  • A data-handling policy
  • A configured Claude workspace

Claude Code security and configuration review

For businesses with development teams

We review how Claude Code is set up across your developers' machines and repositories: permissions, connected tools, secrets, and what it can run without asking. Anything risky is stopped first. Then we fix the configuration and put controls in place so it stays fixed.

You get:

  • Findings ranked by risk
  • A fixed configuration
  • Organisation-wide managed settings

How we work: Contain, Modify, Prevent

  1. Contain

    Before anything else, we find and stop what's risky now: exposed secrets, over-permissive settings and unsafe automations. Every item gets a named owner.

  2. Modify

    We fix the configuration properly and keep what already works, so your team isn't slowed down for no reason.

  3. Prevent

    Organisation-wide controls: managed settings, explicit allow-lists and review rules. Then a handover: a report for leadership, and an engineering report with runbooks.

  • Allow explicitly. Each tool gets a written list of what it may do, not everything minus a few exceptions.
  • Check before you run. Nothing new runs until it has been reviewed.
  • Prove every fix. Each change is tested, and the evidence goes in the report.

Launching soon

ISO/IEC 42001 readiness

ISO/IEC 42001 is the international standard for managing AI responsibly. We're launching a service to take you from your first AI inventory to a certification audit, built around the AI you actually run rather than a stack of templates.

It also gives you a strong head start on the EU AI Act.

Register your interest
  1. Inventory and gap analysis

    Find every place AI is used in your organisation, including the tools staff adopted on their own, and measure it against the standard.

  2. Design your AI management system

    Roles, policy, risk and impact assessment, sized for your organisation.

  3. Controls and evidence

    Put the Annex A controls in place and collect the evidence an auditor will ask for.

  4. Audit readiness

    An internal audit and management review, so the certification audit holds no surprises.

Questions we're often asked

What is an AI security review?

An independent review of your LLM features, agents and AI suppliers, covering prompt injection, data leakage, over-privileged agents and model supply chain risk. You get fixes ranked by impact.

What is prompt injection?

Prompt injection is when text an AI reads, such as an email, web page, document or user message, contains instructions that make it do something it shouldn't, like revealing data or taking an action nobody asked for.

What does a part-time Chief AI Officer do?

Provides senior AI leadership without a full-time hire: setting your AI strategy, deciding where AI is worth using, keeping the risks in view and reporting to your board.

What is ISO/IEC 42001?

The international standard for managing AI responsibly. Our readiness service, launching soon, takes you from your first AI inventory to a certification audit, and gives you a strong head start on the EU AI Act.

What do your cloud security services cover?

Cloud architecture reviews of how your estate is designed, authorised cloud penetration testing, and cloud security posture management (CSPM) that keeps checking your configuration as it changes.

Can you review how our team uses Claude and Claude Code?

Yes. We find which AI tools your staff already use, set a data-handling policy, and review how Claude Code is configured, using our Contain, Modify, Prevent method.

Tell us what you're building

Describe the system, the deadline, and what's worrying you. We'll tell you what we'd look at first.

thearchitect@chillipepper.software