Oman United Kingdom Norway Support portal
Talk to an expert
AboutContact

Prompt injection testing

Adversarial testing of direct and indirect injection paths, including documents and web content, tool abuse and data leakage, with a retest after remediation.

Test an AI application
Standards
ISO/IEC 42001 · NIST AI RMF
Testing
Adversarial, tool-aware

Why this matters now

AI is already in use inside most organisations, usually before governance catches up. Prompt injection testing gives you control without blocking work that is genuinely useful.

What the engagement covers

Each area below is a defined part of the engagement, with an owner, an output and a date.

Scope

A defensible boundary statement agreed before any control work begins.

Attack paths

Direct and indirect injection routes, including documents and web content.

Tool abuse

Attempts to make the system act beyond its intended authority.

Data leakage

Testing for exposure of system prompts, secrets and other users' data.

Mitigations

Specific, implementable fixes ranked by risk reduction.

Retest

Verification after remediation, with a comparison against the first run.

How the work runs

The same sequence on every engagement, so you know what happens next.

STEP 01

Inventory

Establish what AI is actually in use, including tools adopted without approval.

STEP 02

Assess

Evaluate each use case for security, privacy, legal and model risk.

STEP 03

Control

Apply proportionate controls: policy, access, logging, approval points and monitoring.

STEP 04

Test

Adversarial testing against the deployed system, including indirect injection paths.

STEP 05

Govern

Reporting, review and change control so governance keeps up with adoption.

What you get out of it

Adoption without blind spots

Useful AI work continues; the uncontrolled parts get controlled.

Standards alignment

Mapped to ISO/IEC 42001 and ISO/IEC 27001 so one programme serves both.

Tested, not assumed

Controls verified by adversarial testing against the live system.

Standards and references: ISO/IEC 42001:2023 · NIST AI RMF · OWASP Top 10 for LLM Applications · ISO/IEC 27001:2022

Before you enquire

How quickly can this start?
Scoping usually takes one call. Delivery start depends on access and approvals; we will give you a date in the proposal rather than after you sign.
Who actually does the work?
Named consultants from our regional teams. You will meet them before the engagement begins, and the same people write the report.
How is this priced?
A fixed price against a written scope wherever the work can be scoped that way, and a day rate where it genuinely cannot. Either way you see the basis before you commit.

Test an AI application

One scoping call with the consultant who would run the work. No obligation, no charge.

Test an AI application
Talk to an expert