AI risk assessment
Assessment of each use case for security, privacy, legal and model risk, covering what data reaches the model, from where, and under whose agreement.
Assess AI risk- 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. AI risk assessment 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.
Use-case review
Assessment of purpose, data, users and consequence for each use case.
Data
What data reaches the model, from where, and under what agreement.
Model
Model provenance, evaluation, versioning and change control.
Suppliers
Third-party risk assessed proportionately and reviewed on renewal.
Controls
Annex A or framework controls mapped to what you already run.
Treatment plan
Prioritised controls with owners and dates.
How the work runs
The same sequence on every engagement, so you know what happens next.
Inventory
Establish what AI is actually in use, including tools adopted without approval.
Assess
Evaluate each use case for security, privacy, legal and model risk.
Control
Apply proportionate controls: policy, access, logging, approval points and monitoring.
Test
Adversarial testing against the deployed system, including indirect injection paths.
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.
Before you enquire
How quickly can this start?
Who actually does the work?
How is this priced?
Where to go next
Assess AI risk
One scoping call with the consultant who would run the work. No obligation, no charge.