CompTIA SecAI+
CompTIA SecAI+ (CY0-001) is CompTIA's first vendor-neutral AI security certification, launched February 17, 2026. Four domains spanning foundational AI concepts, securing AI systems, AI-assisted security operations, and AI governance/risk/compliance. Maximum 60 multiple-choice and performance-based questions, 60 minutes, passing score 600 on a scaled 100-900 scale. CompTIA recommends Security+, CySA+, or PenTest+ as a prerequisite, though it is not enforced.
Exam fee at last check: $359, per the vendor. The official objectives are at www.comptia.org/en-us/certifications/secai/. Verified 2026-09-26.
Practice is free. The full analysis adds the right answer, the explanation and the source for every question, readiness by domain and a study plan, for one payment.
- Basic AI Concepts Related to Cybersecurity17%
- Securing AI Systems40%
- AI-Assisted Security24%
- AI Governance, Risk, and Compliance19%
Source: Vendor exam objectives. Checked 2026-09-26. Bar length is relative to the heaviest domain; the figure beside each name is the published weight.
What's on the exam
Basic AI Concepts Related to Cybersecurity 17%
Core AI/ML terminology, how AI is used both defensively and by attackers, and the vocabulary the rest of the exam assumes you already have.
- Supervised, unsupervised, and reinforcement learning, and where each shows up in security tooling
- Generative AI fundamentals: LLMs, fine-tuning vs retrieval-augmented generation (RAG)
- AI-driven threats: automated phishing, deepfakes, adversarial machine learning at a conceptual level
- Federated learning and other distributed training architectures
Primary sources: CompTIA SecAI+ (CY0-001) Certification Exam Objectives, NIST AI 100-1, AI Risk Management Framework
Securing AI Systems 40%
The largest domain by far. Protecting AI systems, their data, and their deployment environments from the attack surface that is unique to AI: prompt injection, data poisoning, model extraction, insecure plugins/agents, and supply-chain risk in pretrained models.
- OWASP Top 10 for LLM Applications: prompt injection (direct and indirect), insecure output handling, data/model poisoning, model denial of service, supply chain, sensitive information disclosure, insecure plugin design, excessive agency, overreliance, model theft
- MITRE ATLAS tactics: reconnaissance, ML attack staging, exfiltration, persistence, evasion, impact
- Access control and provenance for model artifacts and training pipelines (signing, hashing, SLSA-style attestation)
- RAG-specific authorization gaps: retrieval must respect the requesting user's actual permissions, not a privileged service account's
Primary sources: OWASP Top 10 for LLM Applications 2025, MITRE ATLAS, NIST AI 100-2e2025, Adversarial Machine Learning: A Taxonomy and Terminology
AI-Assisted Security 24%
Using AI to make security operations faster and better: alert correlation, anomaly-based threat hunting, and generative AI for drafting security content, always with attention to where human review still belongs in the loop.
- Human-in-the-loop vs autonomous automation, and why automating dismissal decisions carries more risk than automating prioritization
- Anomaly-based detection via unsupervised clustering, as a complement to signature-based detection
- Automated alert correlation and enrichment across EDR, network, and cloud telemetry
- Generative AI hallucination risk when used to draft runbooks, reports, or phishing-simulation content
- Evaluating AI-security vendor tools: false-positive rate, false-negative rate, and how those were measured
Primary sources: CompTIA SecAI+ (CY0-001) Certification Exam Objectives, NIST AI 600-1, AI Risk Management Framework: Generative AI Profile
AI Governance, Risk, and Compliance 19%
Regulatory frameworks that govern AI use (EU AI Act, GDPR Article 22), and applying GRC practice (risk tiers, bias/fairness testing, lifecycle risk management) throughout the AI system lifecycle.
- EU AI Act risk tiers: unacceptable, high-risk (includes employment decisions), limited (transparency-only), minimal
- GDPR Article 22: the right not to be subject to a decision based solely on automated processing with legal or similarly significant effect
- NIST AI RMF's four functions: Govern, Map, Measure, Manage
- Bias and fairness testing as part of pre-deployment risk measurement
Primary sources: EU AI Act (Regulation (EU) 2024/1689), GDPR Article 22, NIST AI 100-1, AI Risk Management Framework
Exam-day strategy
This is a short exam (60 minutes, max 60 questions), so pace at roughly 1 minute per question and flag anything that needs a second look rather than stalling on it.
Domain 2 (Securing AI Systems) is 40% of the exam. Know the OWASP LLM Top 10 categories cold and be able to distinguish prompt injection from insecure output handling from excessive agency, they are frequently used as distractors against each other.
When a scenario describes an autonomous agent with broad permissions acting on external content, check for both indirect prompt injection and excessive agency, exam scenarios often test both concepts in the same question.
For governance questions, remember GDPR Article 22 is about automated decisions with legal or similarly significant effect specifically, not GDPR generally, and the EU AI Act's high-risk tier is where most real employment/credit-scoring AI systems land.
Do not confuse model extraction (stealing a model via its API) with data poisoning (corrupting training data) or model inversion (reconstructing training data from outputs). All three show up as distractors for each other.
Additional resources
Explore Related Cybersecurity Resources
DecipherU is independent of CompTIA. SecAI+ is a trademark of CompTIA. DecipherU is not endorsed by, sponsored by, or affiliated with any certifying body.