Artificial intelligence is quickly becoming part of everyday business and cybersecurity operations. While I already work within cybersecurity and have spent a lot of time experimenting with AI personally, I wanted to develop a better understanding of AI from a security, governance and enterprise risk perspective.

That led me to ISACA’s Advanced in AI Security Management (AAISM) certification.

AAISM Certification Overview

The ISACA Advanced in AI Security Management (AAISM) is an advanced certification designed for experienced security professionals responsible for managing the security and risk implications of artificial intelligence within organisations.

Unlike many certifications, AAISM has a prerequisite: candidates must already hold an active CISSP or CISM certification before they can take the exam. The exam contains 90 multiple-choice questions and candidates have 150 minutes to complete it.

To maintain the certification, holders must keep their qualifying CISSP or CISM active, comply with ISACA’s Code of Professional Ethics, and complete 10 AAISM-related CPE hours each year and 30 over a three-year period.

This prerequisite structure positions AAISM as a specialist extension to an existing security-management background rather than an entry-level AI or cybersecurity certification.

Preparing for the exam

My primary study resource was the official ISACA AAISM Review Manual.

I worked through the material systematically, focusing on understanding how ISACA approaches AI governance, risk management, security controls and the wider AI lifecycle rather than simply memorising terminology.

Given my existing cybersecurity background, many of the underlying principles were already familiar. The main challenge was understanding how those established security concepts are extended to AI – particularly around model risk, data governance, responsible AI, third-party dependencies and continuous monitoring.

Overall, I found the material reasonably approachable. The biggest shift was not learning an entirely new security discipline, but learning how existing governance, risk and security practices need to adapt when AI becomes part of an organisation’s technology environment.

Sitting the exam

Going into the exam I felt reasonably confident, although, as with most certification exams, there were still questions where I had to stop and think carefully about what was actually being asked.

I flagged approximately 15 of the 90 questions for review.

After going back through those questions and reviewing my answers, I submitted the exam with roughly an hour remaining.

I was then greeted with the result I had been hoping for:

Preliminary Pass.

That was a pretty good screen to see.

What I actually learned

One thing I appreciated about AAISM was that it wasn’t trying to turn security professionals into data scientists.

Instead, the certification focuses on how AI changes the responsibilities of security professionals.

The material looks at AI as an enterprise security and risk-management problem, covering how organisations govern AI, assess its risks, secure the technology and data behind it, and continue monitoring it after deployment.

The curriculum is divided into three main domains.

AI Governance and Program Management

The first domain covers how organisations establish appropriate governance around AI.

This includes defining ownership, roles and responsibilities; establishing AI strategies, policies and acceptable-use requirements; and making sure AI initiatives remain aligned with wider organisational objectives.

There is also significant emphasis on responsible AI.

Topics include:

• Bias and fairness
• Transparency and explainability
• Trust and safety
• Privacy
• Intellectual property
• Human rights
• Human oversight
• Regulatory and compliance requirements

Another important aspect is simply understanding what AI an organisation is using.

The material covers maintaining inventories of AI systems and associated data, data classification and lineage, security metrics, AI-specific security awareness, and incorporating AI into existing incident-response, business-continuity and disaster-recovery processes.

This domain represents 31% of the AAISM exam.

AI Risk and Opportunity Management

The second domain applies familiar risk-management principles specifically to AI.

This includes identifying AI risk, establishing acceptable risk thresholds, determining appropriate risk responses and carrying out assessments throughout the AI lifecycle.

Frameworks and regulatory approaches covered include the NIST Artificial Intelligence Risk Management Framework and the EU AI Act, alongside concepts such as privacy impact assessments, conformity assessments and fundamental-rights impact assessments.

The more security-focused part of this domain covers the emerging AI threat landscape.

Some of the threats covered include:

• Prompt injection
• Training-data leakage
• Data poisoning
• Model poisoning
• Model theft
• Model inversion
• Model evasion

AAISM also spends time on the less obvious risks surrounding AI.

Organisations increasingly rely on third-party models, cloud services, APIs, open-source components and external datasets. Because of this, vendor management, shared responsibility and AI software supply-chain risk become significant parts of AI security.

This domain also represents 31% of the exam.

AI Technologies and Controls

The third and largest domain provides enough technical background to understand what is actually being protected.

It covers different forms of AI and machine learning, including:

• Generative and predictive models
• Supervised learning
• Unsupervised learning
• Reinforcement learning
• Neural networks
• Agentic AI

It then looks at security architecture and secure-by-design principles for AI systems.

A particularly useful part was viewing AI through its complete lifecycle:

Plan and design → collect and process data → build or adapt models → test and validate → deploy → operate and monitor → retire or decommission.

Security considerations exist throughout every stage.

The material also goes heavily into data governance and security because AI systems are fundamentally dependent on data. That includes data acquisition, storage, retention, destruction, access control, confidentiality, integrity and backups.

Other security concepts include Zero Trust, auditability and traceability, human-in-the-loop controls, shadow AI, supply-chain controls and AI-specific incident management.

Importantly, deploying an AI system isn’t the end of the process.

AI needs continuous monitoring for changes such as model drift, evolving threats, inappropriate use and changes to the underlying models, data or integrations.

This domain represents the remaining 38% of the exam.

Existing cybersecurity principles still matter

Probably my biggest takeaway from studying AAISM was that AI security isn’t an entirely separate discipline from cybersecurity.

A lot of the foundations are already familiar.

What changes is the technology being protected and some of the risks surrounding it.

AI introduces considerations around training data, models, unpredictable outputs, explainability, bias, model drift and attacks that simply didn’t exist in the same form within traditional infrastructure.

The underlying security principles still apply, but they need to be extended to account for those differences.

I think this is an important distinction as organisations increasingly adopt AI. It is very easy for AI security discussions to become focused entirely on the latest models or technical capabilities, when many of the problems organisations will face are ultimately governance, risk and security-management problems.

Was AAISM worthwhile?

For me, yes.

I didn’t pursue AAISM because I expected it to teach me how to train a neural network or build a large language model from scratch.

That isn’t really what the certification is designed for.

Its value is in understanding how AI fits into the wider security environment of an organisation.

As AI adoption continues, security teams are going to have to answer questions such as:

How do we know which AI systems are being used?

What data are employees putting into them?

How do we assess an AI vendor?

Who owns the risk associated with an AI system?

How do we respond if an AI system is compromised?

How do we test and monitor these systems?

What happens when a model begins behaving differently six months after deployment?

And how do we allow an organisation to benefit from AI without simply blocking it because it introduces new risks?

Those are fundamentally security-management questions.

AAISM gave me a much more structured way of thinking about them.

What’s next?

AI security is still developing extremely quickly.

Standards, regulation, attack techniques and the technology itself are all evolving at the same time, which means I don’t see completing AAISM as the end of learning about AI security.

If anything, it has given me a stronger foundation from which to continue exploring the technical side of AI security, AI-enabled cybersecurity and the risks that emerge as organisations integrate increasingly capable AI systems into their environments.

For someone already working in cybersecurity, that was exactly what I wanted to get out of it.