CompTIA SecAI+ · CY0-001 (V1)
If you've decided to sit the SecAI+ exam, you've probably already found the official CompTIA SecAI+ objectives — a 13-page PDF that lists every domain and sub-objective, explains none of them, and (as of document version 2.0) still lists the question count and exam length as "TBD." The blueprint is essential; it just isn't studyable.
This guide fixes that. Below you'll find all four CY0-001 domains and all fifteen sub-objectives explained in plain English — what each one actually means, how it's likely to show up on the exam, and how to turn the weightings into a study order. The exam facts are verified as of July 2026, and we update this page as CompTIA revises the certification.
SecAI+ is CompTIA's first AI-security certification — launched February 17, 2026 — and it certifies that you can secure AI systems, use AI to strengthen security operations, and govern AI responsibly.
| Exam code | CY0-001 (V1) |
| Launched | February 17, 2026 |
| Questions | Up to 60 |
| Time | 60 minutes |
| Question types | Multiple-choice + performance-based (PBQs) |
| Passing score | 600 (scale of 100–900) |
| Recommended experience | 3–4 years in IT, ~2 years hands-on cybersecurity (Security+/CySA+ level) |
| Objectives document | Version 2.0 |
Why some "official" numbers look different: the objectives PDF was finalized before launch, so its "About the Exam" table still shows TBD for several details above. The figures here come from CompTIA's live exam pages and post-launch materials — and it's exactly why a static PDF makes a poor bookmark for a certification this new.
The exam blueprint has four domains, and the weighting tells you something important before you read a single objective: Securing AI Systems alone is 40% of the exam. SecAI+ is first and foremost about defending AI systems — using AI as a security tool comes second, and definitions and governance frame the edges.
Six of the exam's fifteen sub-objectives live in that heavyweight domain, and every one of them begins "Given a scenario…" — CompTIA's signal for applied, often performance-based testing. Plan your study time accordingly; the sections below walk the domains in blueprint order.
If you're coming from Security+, notice how differently this blueprint is shaped. SY0-701 spreads five domains fairly evenly, with nothing above 28%; CY0-001 concentrates 40% in a single domain. That's the profile of a specialist exam: it assumes the broad security foundation and drills deep on one discipline. In practice it means an uneven study plan isn't a mistake here — it's the correct response to the blueprint.
This is the vocabulary domain: the smallest slice of the exam and the fastest place to bank points, but it underpins everything else on the test.
1.1Compare and contrast AI types and techniques used in cybersecurity. Expect definitional and compare-contrast questions across three clusters: types of AI (generative AI, machine learning, deep learning, transformers, NLP — including the difference between large and small language models, and what a GAN is), model training (supervised vs. unsupervised vs. reinforcement learning, fine-tuning, and optimization concepts like pruning and quantization), and prompt engineering (system vs. user prompts, zero-shot vs. one-shot vs. multi-shot prompting, prompt templates). If you can explain why an LLM differs from an SLM and when you'd fine-tune rather than prompt-engineer, you're at the level this objective wants.
1.2Explain the importance of data security in relation to AI. Models are only as trustworthy as their data, so this objective covers data hygiene concepts — lineage, provenance, integrity, cleansing — plus structured vs. unstructured data, watermarking, and the components of retrieval-augmented generation (RAG), including vector storage and embeddings. RAG is a strong candidate for exam questions because it's where enterprise data meets model behavior.
1.3Explain the importance of security throughout the AI life cycle. This walks the model development life cycle from business case through data collection, model development and evaluation, deployment, and ongoing monitoring — with a security lens at each stage. Know the human-centric design principles by name: human-in-the-loop, human oversight, and human validation. Life-cycle-ordering questions ("at which stage should X happen?") are classic CompTIA territory.
Two-fifths of the exam and six sub-objectives, every one scenario-based. This is where the pass/fail margin lives, and it's the domain to study first and revisit most.
2.1Use AI threat-modeling resources. You're expected to know the industry's AI threat catalogs and when to reach for each: the OWASP LLM Top 10 and ML Security Top 10, MITRE ATLAS (the adversarial-AI counterpart to ATT&CK), the MIT AI Risk Repository, and the CVE AI Working Group. A likely question shape: a described weakness, and you pick the framework or entry that models it.
2.2Implement security controls for AI systems. The control families here split into model controls (guardrails, model evaluation, prompt templates) and gateway controls — the chokepoint in front of the model: prompt firewalls, rate limits, token limits, input quotas, modality limits, and endpoint access controls. Expect requirement-matching questions: given a set of requirements, choose the control stack that satisfies them.
2.3Implement appropriate access controls for AI systems. Short but consequential: access decisions now exist at four layers — model access, data access, agent access, and network/API access. The new one for most candidates is agent access: what an autonomous AI agent is allowed to touch is an exam-relevant design decision.
2.4Implement data security controls for AI systems. Encryption in transit, at rest, and — the differentiator — in use, plus the data-safety toolkit: anonymization, classification labels, redaction, masking, and minimization. Scenario questions will hinge on picking the right technique for a described data exposure.
2.5Implement monitoring and auditing for AI systems. AI observability goes beyond logs: prompt monitoring (both queries and responses), log sanitization and protection, response-confidence tracking, rate monitoring, and — uniquely — AI cost monitoring across prompts, storage, and processing. Auditing extends to quality and compliance: hallucinations, accuracy, bias and fairness, and access.
2.6Analyze evidence of an attack and suggest compensating controls. The blueprint's biggest single objective and prime PBQ material. Know the AI attack taxonomy — prompt injection, jailbreaking, data and model poisoning, model inversion, model theft, membership inference, AI supply-chain attacks, model denial of service, insecure output handling, excessive agency, and more — well enough to recognize each from described evidence, then map it to compensating controls like prompt firewalls, guardrails, least privilege, rate limiting, and data-integrity controls. If you drill one thing from this whole page, drill attack-to-control mapping.
The defender's-toolkit domain: using AI to do security work, and understanding how attackers use it too. It's also the domain most likely to overlap with your day job — if you already use AI assistants in a SOC or pipeline, you're partway there.
3.1Use AI-enabled tools to facilitate security tasks. Covers the tool landscape (IDE, browser, and CLI plug-ins; chatbots; personal assistants; and — notably current — Model Context Protocol servers) and the use cases the exam cares about: anomaly detection, vulnerability analysis, automated penetration testing, incident management, threat modeling, fraud detection, and summarization. Questions will pair a task with the right tool category.
3.2Explain how AI enables or enhances attack vectors. The offense side: deepfakes for impersonation and disinformation, AI-accelerated reconnaissance and social engineering, obfuscation, automated data correlation, and automated attack generation — from payloads and malware to DDoS. This objective is conceptual ("explain"), so expect recognition rather than hands-on simulation.
3.3Use AI to automate security tasks. Automation in practice: low-code/no-code scripting, document synthesis, incident-ticket management, AI-assisted change management (approvals, automated deployment and rollback), AI agents, and AI in CI/CD pipelines — code scanning, software composition analysis, unit and regression testing, and model testing. If you've never watched an AI agent triage a ticket queue or a pipeline gate on a model test, review this area with extra care.
Don't let "GRC" read as dry-policy trivia. Candidates who've sat the exam consistently report these questions arrive as applied scenarios — a company adopting AI, and you choose the governance response — not clause recall.
4.1Explain organizational governance structures that support AI. Two halves: structures (an AI Center of Excellence, AI policies and procedures) and the expanding cast of AI-related roles — data scientist, AI architect, ML engineer, MLOps engineer, AI security architect, AI governance engineer, AI risk analyst, AI auditor. Be ready to match a responsibility to the right role.
4.2Explain risks associated with AI. Anchored by the responsible-AI principles — fairness, reliability and safety, transparency, privacy, explainability, inclusiveness, accountability — plus the practical risk list: bias introduction, accidental data leakage, reputational damage, model accuracy drift, IP risks, autonomous-system risk, and shadow AI (unsanctioned AI use inside the business — the modern shadow IT, and an easy scenario to write questions about).
4.3Summarize the impact of compliance on AI use and development. The regulatory map: the EU AI Act, NIST AI Risk Management Framework (AI RMF), ISO AI standards, and OECD standards, plus corporate-policy dimensions — sanctioned vs. unsanctioned tools, private vs. public models, sensitive-data governance, third-party compliance evaluations, and data sovereignty. You don't need to memorize statutes; you need to know which framework applies to which situation.
The blueprint above is also a ready-made study sequence — you just have to follow the weights instead of the page order.
It's 40% of the exam, six of the fifteen objectives, and the home of the scenario/PBQ questions. Attack-to-control mapping (2.6) is the single highest-value skill on the test.
Together they're another 43%. Domain 3 rewards hands-on familiarity with AI tooling; Domain 4 rewards thinking in scenarios — practice choosing frameworks and roles for described situations rather than memorizing definitions.
It's vocabulary. Review it early for orientation, then reinforce with flashcards throughout rather than dedicating a study block. CompTIA also publishes an acronym list with the objectives — from ATLAS to RAG to MCP — that makes an ideal flashcard deck.
Rate yourself 1–3 against each of the fifteen objectives above. Anything rated 1 gets targeted practice before your next full review pass. When your practice scores are consistently clearing the passing threshold with margin, book the exam.
Practice against this exact blueprint: the FlashGenius SecAI+ Study Hub gives you interactive, domain-organized prep for CY0-001, and the SecAI+ Ultimate 2026 Guide covers the career context — salary, roles, and how SecAI+ fits alongside Security+ and CySA+.
Download the official blueprint from the CompTIA SecAI+ certification page — it's the primary source, and this article annotates it rather than replacing it. Just know its limits: it's a static, pre-launch document (version 2.0 still carries TBDs), and it lists what is tested with no explanation of what any of it means. This page tracks the live facts and gets updated as CompTIA revises the exam.
One more thing the official PDF makes explicit: CompTIA's authorized-materials policy states that candidates who prepare with unauthorized "braindump" content have their certifications revoked and are suspended from future testing. That policy is printed inside the objectives document itself. If you're weighing shortcut sites for any CompTIA exam, read our breakdown of why exam dumps are a bad bet — the risk calculus is the same here, on a certification too new for any dump to even pretend to be current.
Up to 60 questions in 60 minutes, mixing multiple-choice and performance-based questions. The passing score is 600 on a scale of 100 to 900. Note that the official objectives PDF still lists some of these details as TBD because it was finalized before the February 2026 launch.
Securing AI Systems (40%), AI-assisted Security (24%), AI Governance, Risk, and Compliance (19%), and Basic AI Concepts Related to Cybersecurity (17%) — fifteen sub-objectives in total across the four domains.
There's no formal prerequisite, but CompTIA recommends 3–4 years of IT experience including about 2 years of hands-on cybersecurity — roughly Security+ or CySA+ level knowledge. SecAI+ assumes you already understand core security concepts and adds the AI layer on top.
Yes. The exam combines multiple-choice and performance-based questions, and the blueprint signals where they concentrate: every objective in the 40% Securing AI Systems domain begins "Given a scenario," which is CompTIA's marker for applied testing.
It depends on your baseline. CompTIA estimates 20–25 hours of structured learning for experienced security professionals; if AI security is new to you, plan more like 4–6 weeks part-time, weighted toward the Securing AI Systems domain and hands-on familiarity with AI tools.
It's different rather than strictly harder: a shorter exam (60 questions vs. 90) with a narrower but newer body of knowledge, and it assumes the security fundamentals Security+ tests. Candidates comfortable with core security but new to AI concepts usually find Domains 1 and 2 carry the learning curve.
The objectives are the exam. Everything CompTIA can ask on CY0-001 maps to the fifteen sub-objectives explained above — so study the blueprint, in weight order, and self-score as you go: Securing AI Systems first and deepest, AI-assisted Security and GRC next, vocabulary throughout. When you're ready to test yourself against the blueprint, start with the SecAI+ Study Hub — and check back here as CompTIA updates the exam, because on a certification this new, currency is half the battle.
