AI application assessment
Examine the model integration, retrieval flow and application permissions against relevant misuse scenarios.
All services
Service portfolio / 07
AI & emerging technology security
Understand the new trust boundaries created when models, tools and business data are connected.
Explore the scopeService scope
The work.
The context.
The next step.
AI systems introduce decisions about what a model can see, which tools it can use and who checks its output. Assess those boundaries in the context of the actual application and its intended use.
Prompt, tool & data controls
Review untrusted input handling, sensitive-data exposure and the permissions granted to connected tools.
AI governance & supplier review
Define ownership, evaluation criteria and review points for model and provider changes.
Secure adoption architecture
Plan access boundaries, human approval points and operational logging for new AI-enabled workflows.
When this is useful
A starting point
that fits your need.
- An internal AI assistant or agent rollout
- Sensitive data connected to a model
- Unclear approval and oversight requirements
Deliverables to define in your scope
AI system boundary and risk assessment
Prioritised architecture and application controls
Evaluation scenarios and governance recommendations
Connected expertise
Build on the findings.
Connect this engagement to the other parts of your security programme when the scope calls for it.
Scope the work around your business.