Overview/Register/ Secure AI Architecture & Engineering/AE-02
AI Threat Modeling
Finds how attackers can misuse or attack an AI system, early in the design.
AE-02 What this service does
Finds how attackers can misuse or attack an AI system, early in the design.
It is delivered inside the Secure AI Architecture & Engineering practice area, which exists to design the security of AI systems before they are built, and put guardrails on what already runs.
The engagement is scoped at the briefing; nothing here is priced on the page. What is fixed is the output: the artefacts below are what you hold at the end.
AE-02-D Deliverables
On handover- Threat model
- abuse cases
- control recommendations
Where it sits Delivery context
- Practice area
- Secure AI Architecture & Engineering (Tab F, reference AE)
- Service reference
- AE-02
- Indicative tier
- High. Tiers are indicative of the risk this service addresses and are not client data.
- Why the area exists
- Design the security of AI systems before they are built, and put guardrails on what already runs.
- Evidence
- Practitioner credentials behind this work are listed once, in full, on the credentials page. We publish no client names.
Same tab Other services in Secure AI Architecture & Engineering
-
AE-01
Secure AI Architecture Design & Review
Designs or reviews the security architecture of AI systems: model hosting, data stores, APIs and integrations.
-
AE-03
Secure AI SDLC & MLOps Security
Adds security requirements, reviews and automatic tests to how teams build, train and deploy AI.
-
AE-04
AI Guardrail Design & Implementation
Designs and installs runtime guardrails that block prompt injection, unsafe output and data leaks.
We are not accepting new clients at present, so no briefing is running on AE-02. What this service produces is set out above.
Read the availability note