CY0-001 Domains Explained

The short version

CY0-001 has four domains: basic AI concepts (17%), securing AI systems (40%), AI-assisted security (24%), and AI governance, risk, and compliance (19%). The weightings come from CompTIA’s SecAI+ page, checked October 18, 2026. The topic lists below follow CompTIA’s exam objectives document, version 2.0, which we read; CompTIA says its example lists are not exhaustive, so always check the current objectives before you rely on any list.

Domain 1: Basic AI concepts related to cybersecurity (17%)

What it covers. Three ideas. First, the types of AI and how they are trained: generative AI, machine learning, deep learning, language models (large and small), supervised, unsupervised, and reinforcement learning, fine-tuning, and prompt techniques such as zero-shot, one-shot, and multi-shot prompting. Second, data security: cleansing, verification, lineage, provenance, integrity, balancing, augmentation, watermarking, and retrieval-augmented generation (RAG) with embeddings and vector storage. Third, security across the AI life cycle, from business case through data collection, model development, evaluation, deployment, validation, and monitoring, including human-in-the-loop, oversight, and validation.

In practice. You will meet short scenarios: a dataset with duplicates, a model that needs a smaller footprint, or a regulator that wants a person to approve each output.

Typical traps.

  • Mixing up human-in-the-loop (a person inside the process), human oversight (a person supervising the system), and human validation (a person checking an output).
  • Treating data augmentation as always safe. If the requirement says every record must be a real observation, synthetic additions are out.
  • Confusing fine-tuning with prompt engineering.

Practice. Build a one-page table of terms with a single sentence and one security concern each.

Domain 2: Securing AI systems (40%)

What it covers. CompTIA breaks it into six objectives:

  • Threat-modeling resources, such as the OWASP Top 10 lists for LLMs and machine learning, the MIT AI Risk Repository, and MITRE ATLAS.
  • Security controls: model guardrails and prompt templates, and gateway controls such as prompt firewalls, rate limits, token limits, input quotas, and endpoint access.
  • Access controls for models, data, agents, and APIs.
  • Data security: encryption in transit, at rest, and in use, plus anonymization, classification labels, redaction, masking, and minimization.
  • Monitoring and auditing: prompts and responses, logs and log sanitization, confidence, rate and cost, and audits for hallucination, accuracy, bias, and access.
  • Analyzing evidence of attacks and choosing compensating controls.

In practice. Scenarios ask what you would add or change given a constraint, such as a vendor-hosted model you cannot retrain or a latency limit.

Typical traps.

  • Choosing a control that is real but aimed at a different layer. A rate limit bounds volume and does not stop a few cleverly worded messages.
  • Mixing up masking, redaction, and anonymization.
  • Confusing data poisoning (bad training data) with model poisoning (a tampered model), or model inversion (recovering training data) with membership inference (testing whether a record was in the training set).

Practice. Make a two-column list of every attack in the objectives and the compensating control that answers it. Then rehearse it from memory. See the study plan for where this fits.

Domain 3: AI-assisted security (24%)

What it covers. Using AI-enabled tools for security tasks (IDE, browser, and command-line plug-ins, chatbots, assistants, and Model Context Protocol servers), with use cases such as signature matching, vulnerability analysis, anomaly detection, pattern recognition, fraud detection, and summarization. Then how AI helps attackers: deepfakes and impersonation, misinformation and disinformation, reconnaissance, social engineering, obfuscation, and automated attack generation. Finally, automating security work with low-code and no-code scripting, AI agents, ticket handling, change management, and CI/CD stages such as code scanning, software composition analysis, regression testing, and model testing.

In practice. Questions describe an incident or a pipeline and ask which AI capability or stage fits.

Typical traps.

  • Misinformation (false, not necessarily deliberate) versus disinformation (deliberately false).
  • Anomaly detection (departures from normal) versus pattern recognition (matching known shapes).
  • Forgetting that a pipeline built for code changes will not notice a change in the model. Model testing is its own stage.

Practice. For each use case, write one example of where it works well and one limit.

Domain 4: AI governance, risk, and compliance (19%)

What it covers. Governance structures such as an AI Center of Excellence and AI policies, and the roles around AI (data scientist, ML engineer, MLOps engineer, AI auditor, and others). Responsible-AI principles including fairness, transparency, explainability, accountability, privacy, and inclusiveness. Risks such as bias, data leakage, intellectual property issues, and shadow AI. Compliance frameworks: the EU AI Act, OECD standards, ISO AI standards, the NIST AI Risk Management Framework, data sovereignty, and corporate policies.

In practice. You match a requirement to a principle, a role, or a framework function.

Typical traps.

  • Transparency (disclosing that AI is used) versus explainability (giving the reason for one decision).
  • Policy (the rule) versus procedure (the steps).
  • Assuming a framework like NIST’s replaces your existing risk process. It is meant to fit into it.

Practice. Write one sentence for each principle and one example of a requirement that tests it.

More guides for CY0-001

Next step

Try the sample questions to see these domains in scenario form. Our CY0-001 practice tests report results per domain, so you can see which of the four needs work.

Try these practice exams