
CompTIA SecAI+ (CY0-001) is CompTIA's certification for AI security, launched on 17 February 2026. It covers two sides of the same problem: securing AI systems (models, data pipelines, prompts and inference endpoints), and using AI to make security operations faster. It also covers AI governance, risk and compliance, including GDPR and the NIST AI Risk Management Framework. Securing AI systems carries 40% of the exam. It suits SOC analysts, security engineers and cloud security engineers whose companies are rolling out GenAI and AI agents.
This guide follows CompTIA's official SecAI+ certification page: exam facts, the four domains with weights, the main AI threats and defences, a 6-week plan and 15 original practice questions.
Quick facts (CompTIA, November 2026)
| Item | Detail |
|---|---|
| Exam code | CY0-001 |
| Launched | 17 February 2026 |
| Questions | Maximum of 60, multiple-choice and performance-based |
| Duration | 60 minutes |
| Passing score | 600 on a 100–900 scale |
| Languages | English and Japanese |
| Recommended experience | 3–4 years in IT and 2+ years of hands-on cybersecurity; Security+, CySA+, PenTest+ or equivalent |
| Price | Check CompTIA's store for India pricing |
Read the official CompTIA SecAI+ page and download the exam objectives before you book.
The four domains and weights
| Domain | Weight | What it covers (from CompTIA) |
|---|---|---|
| 1. Basic AI concepts related to cybersecurity | 17% | Machine learning, deep learning, NLP and automation; AI uses in threat detection and security operations; AI-driven threats such as automated phishing, polymorphic malware, adversarial ML and misuse of generative AI |
| 2. Securing AI systems | 40% | Controls that protect AI systems, data and models; securing on-premises, cloud and hybrid deployments; reducing adversarial risks to models, data pipelines and inference layers |
| 3. AI-assisted security | 24% | AI for detection and response, including anomaly detection and faster remediation; automating triage, alert correlation and response orchestration; AI in threat modelling, behaviour analysis and continuous monitoring |
| 4. AI governance, risk, and compliance | 19% | Global regulations and their effect on AI adoption; GRC across the AI lifecycle; responsible AI, ethics, legal standards, GDPR and the NIST AI RMF |
AI threats and the defences that match
| Threat | What happens | Defences |
|---|---|---|
| Prompt injection (direct and indirect) | Crafted input, or text hidden in a web page or document the model reads, overrides the system's instructions | Treat model output as untrusted; separate instructions from data; least-privilege tools for agents; human approval for risky actions; input and output filtering |
| Sensitive data disclosure | The model reveals personal data, secrets or internal documents | Data minimisation, masking before training or retrieval, access-aware retrieval (RAG that checks permissions), output filters |
| Data poisoning | Attackers tamper with training or retrieval data to change model behaviour | Data provenance, integrity checks, controlled data pipelines, evaluation before release |
| Evasion (adversarial examples) | Inputs are slightly altered so a model misclassifies them | Adversarial testing, robust training, ensembles and monitoring |
| Model theft and extraction | Attackers copy model weights or rebuild the model through many queries | Access control, encryption, rate limits, query monitoring |
| Supply chain risk | Malicious or vulnerable models, datasets or libraries | Trusted registries, scanning model files, SBOMs and signing |
| Excessive agency | An AI agent with broad permissions takes harmful actions | Scoped credentials, allow-lists of tools, approval steps and full logging |
Useful references: the OWASP Top 10 for LLM Applications, MITRE ATLAS (a knowledge base of adversary tactics against AI systems) and the NIST AI RMF.
6-week study plan
| Week | Focus | Hands-on |
|---|---|---|
| 1 | Domain 1: ML, deep learning, NLP, LLMs, RAG and agents; AI-driven attacks | Build a small RAG chatbot with a free-tier model |
| 2–3 | Domain 2: securing data, models, prompts, APIs and deployments; adversarial risks | Test your own chatbot against the OWASP LLM risks; add input and output filters and logging |
| 4 | Domain 3: AI in SIEM and SOAR, anomaly detection, triage and correlation | Use an AI assistant to summarise alerts from sample logs; check every answer against the raw data |
| 5 | Domain 4: regulations, NIST AI RMF, GDPR, India's DPDP Act, 2023, responsible AI | Write a one-page AI use policy and a risk register for the chatbot |
| 6 | Review and timed practice | Book the exam |
Only test AI systems you built or have written permission to test.
15 practice questions with answers
These are original questions written for this guide, not exam questions.
1. What is indirect prompt injection?
Malicious instructions hidden in content the model reads, such as a web page, email or document, rather than typed by the user.
2. Why should LLM output be treated as untrusted input?
It can contain injected instructions, code or links, so passing it straight to tools, browsers or databases is dangerous.
3. What is data poisoning?
Tampering with training or retrieval data to change a model's behaviour.
4. What is an adversarial example?
An input changed slightly so a model makes a wrong prediction, often invisible to people.
5. How do you reduce the risk of an AI agent deleting production data?
Give it least-privilege credentials, an allow-list of tools and a human approval step for destructive actions.
6. Which control helps stop a RAG chatbot from showing documents a user can't access?
Permission-aware retrieval that filters results by the user's access rights.
7. Which technique detects unusual login behaviour without labelled attack data?
Unsupervised anomaly detection.
8. Name one way AI speeds up a SOC.
Automated triage and correlation of related alerts into one incident.
9. Why must analysts verify AI-generated incident summaries?
Models can hallucinate or miss evidence, so decisions must rest on the underlying data.
10. What is model extraction?
Rebuilding a copy of a model by sending many queries and studying its outputs.
11. Which framework from NIST guides AI risk management?
The NIST AI Risk Management Framework (AI RMF).
12. Which knowledge base lists adversary tactics against AI systems?
MITRE ATLAS.
13. How do you protect against malicious model files from public hubs?
Use trusted sources, scan and verify files, prefer safe formats and track them in an SBOM.
14. Which SecAI+ domain has the most weight?
Securing AI systems at 40%.
15. What is the SecAI+ passing score?
600 on a 100–900 scale.
Before and after SecAI+
- Build the base with Security+, CySA+ or PenTest+.
- Add cloud AI security with SC-500 and read AI for cloud security engineers.
- Learn hands-on in our Cyber Security course and AI, GenAI and Agentic AI course; see live security jobs.
Cloud Soft Solutions in Ameerpet, Hyderabad teaches AI, cloud security and cyber security with real-time projects. See the APEX program and all courses.
📞 Call +91 96660 19191 · WhatsApp · Flat No. 513, 5th Floor, Nilagiri Block, Aditya Enclave, Ameerpet, Hyderabad
Frequently Asked Questions
What is CompTIA SecAI+?
CompTIA's AI security certification (exam CY0-001), launched on 17 February 2026, covering securing AI systems, AI-assisted security and AI governance.
How long is the SecAI+ exam?
60 minutes, with a maximum of 60 multiple-choice and performance-based questions.
What is the SecAI+ passing score?
600 on a scale of 100–900.
Which SecAI+ domain has the most weight?
Securing AI systems at 40%.
What experience does CompTIA recommend for SecAI+?
3–4 years in IT and 2+ years of hands-on cybersecurity, with Security+, CySA+, PenTest+ or equivalent.
Which languages is SecAI+ offered in?
English and Japanese.



