ISACA Advanced in AI Security Management (AAISM)

Learning plan

About 40 hours of focused study, split into 4 phases. Pick your weekly pace and the timeline adjusts.

Estimated finish: 4 weeks

What the exam covers

Source: official exam guide. Weights can change, so check before you start.

Before you start

• Eligibility: you need an active CISM or CISSP to register. AAISM assumes you already know security programme management
• Basic familiarity with how LLM applications are built (prompts, RAG, fine-tuning, APIs) so the controls domain makes sense

1. Weeks 1–2 · 11 h

   AI security governance and programme management (Domain 1)

   • Map stakeholder, framework and regulatory requirements (EU AI Act, NIST AI RMF, ISO/IEC 42001) to an AI security charter
   • Write AI-specific strategy, policy and procedures, including acceptable use, model approval and shadow-AI handling
   • Manage the AI asset and data life cycle, with an inventory of models, datasets, prompts and agents and their owners
   • Extend business continuity and incident response to AI failures such as model outage, harmful output and data leakage through prompts

   Hands-on: Write an AI incident response playbook for a customer-facing LLM chatbot that leaked another customer's data. Include detection signals, severity levels, containment (kill switch, prompt rollback), notification triggers and lessons learned.

   Use: AAISM exam content outline · ISACA AI resources

2. Weeks 2–3 · 11 h

   AI risk management (Domain 2)

   • Run an AI risk assessment with explicit thresholds and choose a treatment (mitigate, transfer, avoid, accept)
   • Maintain an AI threat and vulnerability register using MITRE ATLAS techniques and the OWASP LLM Top 10
   • Assess AI vendors and the supply chain (model provenance, third-party APIs, open-weight model risks, contract clauses on data use)

   Hands-on: Build an AI risk register for a RAG-based internal knowledge assistant. List at least eight risks (prompt injection, sensitive-data exposure, poisoned documents, vendor outage and others), each with likelihood, impact, owner, the related ATLAS/OWASP ID and a treatment decision.

   Use: MITRE ATLAS · OWASP Top 10 for LLM Applications · NIST AI 600-1 Generative AI Profile

3. Weeks 3–4 · 12 h

   AI technologies and controls, the largest domain (Domain 3)

   • Design AI security architecture with input/output filtering, guardrails, least-privilege tool access for agents, secrets isolation and network segmentation for model endpoints
   • Secure the model life cycle, covering model selection, training-data integrity, validation, signing and registry controls
   • Apply data controls (classification, masking, retention) and privacy, ethics, trust and safety controls
   • Define monitoring for AI systems, such as prompt/response logging, abuse detection, drift alerts and red-team cadence

   Hands-on: Play Gandalf to level 5 or beyond, writing down each bypass technique. Then map each one to the control that would have stopped it (output filtering, system-prompt isolation, canary tokens) and add those controls to your Phase 2 risk register.

   Use: OWASP Top 10 for LLM Applications · Gandalf prompt-injection challenge (Lakera)

4. Week 4 · 6 h

   Practice questions and review

   • Complete the free 12-question AAISM practice quiz, then at least 150 QAE database questions
   • Re-study any subtopic (1A–3E) where you score below 70%
   • Practise choosing the manager's answer (governance, risk-based priority) over the hands-on technical fix

   Use: Free AAISM practice quiz · ISACA exam candidate guides

You're ready to book when…

• You can explain prompt injection, data poisoning and model extraction, and name a preventive and a detective control for each
• You have a finished AI risk register and AI incident playbook you could present to a steering committee
• You consistently score 75% or higher on unseen QAE questions

Exam tips

• As with CISM, the best answer is usually the one that aligns with business objectives and risk appetite, not the most technical fix.
• Registration gives you a six-month window to test, and you have five years after passing to apply for the credential ($50 fee).