Hello, you are using an old browser that's unsafe and no longer supported. Please consider updating your browser to a newer version, or downloading a modern browser.
AI+ Security Expert is the middle tier of AI CERTs' three-level AI security track, tested by the 50-question AT-2102 exam. This guide covers the 12 modules and their weights, the one-year renewal cycle, what AI CERTs claims about accreditation, and how it compares to SecAI+, COASP and AAISM.
AI+ Security Expert, formerly AI+ Security Level 2, is AI CERTs' intermediate AI security certification, tested by exam AT-2102.
It covers two jobs that more and more often land on the same person: securing AI applications, and using AI and automation in security work. The 12-module blueprint runs from threat modeling an LLM application to prompt injection testing, RAG security, ML pipeline controls, secure cloud deployment, AI incident response, governance and red teaming. Python automation has a module of its own, worth 9 percent of the exam, so this is not a vocabulary test.
AI CERTs is a New York-based certification body. Its numbered AI security track has three tiers (Practitioner, Expert and Strategist), alongside separate compliance, network and ethical hacking credentials. Expert has no mandatory prerequisites, and AI CERTs does not require AI+ Security Practitioner first. The certification is issued by AI CERTs under exam code AT-2102 and is valid for one year, after which you renew it by passing a recertification exam.
Modules 1 to 3. AI security scope, threat modeling and Python automation. 23% combined.
Modules 4 to 6. Application controls, LLM attacks and retrieval security. 27% combined.
Modules 7 and 8. Model and data integrity, MLSecOps, cloud, API and identity. 18% combined.
Modules 9 to 12. Monitoring and response, governance, red teaming and the capstone. The largest group at 32%.
Four things that set AI+ Security Expert apart from the other AI security credentials on the market in 2026.
AI CERTs' numbered AI security track has three tiers: AT-2101 Practitioner, AT-2102 Expert and AT-2103 Strategist. Expert is the intermediate tier. Despite the name, AI CERTs' certification scheme says it validates intermediate-level knowledge and does not imply expert-level competency. The same scheme says the level designation reflects depth of content, so you can start here without Practitioner.
Most of the blueprint sits on the AI application itself: GenAI controls, prompt injection, RAG pipelines, model and data integrity, and the cloud and identity layer underneath. The course material references the OWASP LLM Top 10, MITRE ATLAS, the NIST AI RMF and ISO/IEC 42001.
AI CERTs certifications last one year, and the only way to renew is to pass a recertification exam. That keeps the credential current, at the cost of a yearly fee and sitting. Our rundown of every major AI certification in 2026 shows where AI CERTs sits among the other AI credentials.
AI CERTs says it operates as an independent certification body under ISO/IEC 17024:2012, and its accreditation page says it is actively pursuing ANAB accreditation. It does not claim a DoD 8140 listing for this certification, so do not count on it for a DoD cyber position requirement. The current matrix is published at the DoD Cyber Exchange.
Its practical value is the content. The blueprint is narrow and current, aimed at the parts of an AI system that security teams are now asked to defend, and it fits in a 90-minute online exam.
The certification, the AT-2102 exam, and how to keep AI+ Security Expert current, as AI CERTs published them in October 2026.
Expert sits in the middle of AI CERTs' track. From here, most people either move up into strategy and governance or go deeper into offensive testing. The three options below are the common next steps, one from AI CERTs and two from other certification bodies.
Exam AT-2103, the top tier of AI CERTs' track. AI CERTs describes it as validating advanced-level expertise in AI-driven security strategy, governance and risk management, with questions on security architecture, AI-enabled threat intelligence and enterprise decision-making.
For people who run an AI security program rather than build the controls. AAISM requires an active CISM or CISSP and has a 90-question, 150-minute exam, which makes it a manager's credential more than an engineer's.
For those who want to break AI systems for a living. The COASP exam, 312-52, runs six hours and mixes 65 multiple-choice questions with five practical ones. That is a big step up in hands-on testing from AT-2102, which is all multiple choice.
Expert does not require Practitioner, but most people arrive with some security grounding first. After it, the path splits by whether you want to set strategy or test and defend systems.
The security groundwork
The AI security credential
Strategy or hands-on testing
Two questions to answer before you commit: can you sit it, and should you pursue AI+ Security Expert specifically. Here's a straight answer to both.
No prerequisites, and Practitioner is not required.
AI CERTs' certification scheme sets no mandatory education, experience or certification requirements. You agree to the candidate policies, pay the fee, and test online through AI CERTs' remote proctoring from your own computer, with a webcam, a single screen and a quiet room.
The content is technical from the first module.
AI CERTs recommends a grounding in cybersecurity concepts, AI and machine learning fundamentals, cloud and application security, security assessment methods, and programming. Module 3 is Python automation, and the GenAI and RAG modules expect you to follow how an LLM application is built. Without that base, plan extra study time.
AI CERTs links AI+ Security Expert to the six roles below. In each one, the certification adds the AI-specific layer: securing the AI applications, data and pipelines the job now touches.
Uses AI techniques for threat detection, security analysis and risk identification, and assesses the AI applications themselves. Threat modeling, prompt injection testing and RAG security are the core modules for this role.
Implements security controls and AI-powered defenses. The GenAI controls, ML pipeline and secure deployment modules map to the engineer who builds the guardrails around models and APIs.
Monitors, investigates and responds to incidents with AI tools. The monitoring and incident response module covers prompt logs, detection rules for prompt injection and data exfiltration, and containment steps for AI systems.
Protects the cloud services AI workloads run on. Module 8 covers model endpoints, API gateways, workload identity, key rotation, tenant isolation and network isolation for AI deployments.
Finds weaknesses before attackers do. The red teaming and prompt injection modules cover AI-specific test methods, rules of engagement and how to report findings in business terms.
Advises organizations on AI security strategy and controls. The governance module adds the NIST AI RMF, ISO/IEC 42001 and EU AI Act high-risk requirements to the technical content.
Four AI security credentials that people often weigh against each other. Shortest version: AI+ Security Expert and SecAI+ cover hands-on defense, COASP covers offensive testing, and AAISM covers program management.
| AI+ Security Expert | CompTIA SecAI+ | EC-Council COASP | ISACA AAISM | |
|---|---|---|---|---|
| Issuer | AI CERTs | CompTIA | EC-Council | ISACA |
| Exam | AT-2102 | CY0-001 | 312-52 (beta) | AAISM |
| Questions and Time | 50 multiple choice, 90 min | Up to 60, 60 min | 65 multiple choice plus 5 practical, 6 hours | 90 questions, 150 min |
| Passing Score | 70% (35 of 50) | 600 on a 100 to 900 scale | 70% to 80% by exam form | 450 out of 800 |
| Delivery | Online, AI CERTs proctoring | Pearson VUE or OnVUE online | ECC Exam Portal, remote proctored | PSI, in person or remote |
| Prerequisites | None | None (Security+, CySA+ rec.) | 2 yrs infosec experience or official training | Active CISM or CISSP |
| Validity and Renewal | 1 year, renew by exam | 3 years, continuing education | 120 ECE credits over 3 years | 30 CPEs over 3 years |
| Focus | Securing AI apps, LLMs, RAG and pipelines | Hands-on AI security controls | Offensive testing of AI systems | AI security program management |
| Best For | Engineers and analysts securing AI applications | Practitioners on the CompTIA path | AI red teamers and pen testers | Managers running AI security |
Facts for SecAI+, COASP and AAISM come from our guides to each. Those guides report no DoD 8140 listing for any of the three, and AI CERTs claims none for AI+ Security Expert. Pick by the work you do and by what your employer recognizes.
Certification Matchups: the published facts, strengths and weak spots of 33 certifications, side by side.
Our AI+ Security Expert boot camp is five days of live, instructor-led training, in person or online, built on AI CERTs' authorized courseware with videos, e-books and quizzes. It includes hands-on labs and a capstone project, the exam voucher, an exam retake, and 180 days of one-on-one coaching after class.
Where AI CERTs sits in the wider AI certification field, the alternatives from CompTIA, EC-Council and ISACA, and the OWASP agentic AI risks that overlap the prompt injection and red teaming modules.
The cert-by-cert rundown of the AI credentials employers ask about, with AI CERTs' AI+ Security and AI+ Ethical Hacker listed alongside the CompTIA, ISACA, IAPP and EC-Council options.
A look at AI CERTs' AI+ Ethical Hacker, what it covers, who it suits, and how it differs from CEH. Worth reading if the offensive side of the AI CERTs catalog interests you.
The ten agentic AI risks, ASI01 to ASI10, from the OWASP GenAI Security Project, and why agents that call tools and hold credentials need a list of their own.
Which AI credentials are pulling ahead, judged by institutional credibility, regulatory pressure and what employers are writing into job descriptions.
EC-Council's Enterprise AI Credential Suite, including the offensive COASP compared on this page, and the role each of the four credentials is built for.
A straight look at CompTIA's AI security credential, the closest vendor-neutral alternative to AI+ Security Expert, and at who should hold off for now.
Why SecAI+, AAISM and AAIA suit practitioners, managers and auditors in turn, and how to match the credential to the job you actually do.
AI CERTs' current blueprint splits AT-2102 into 12 modules, and no single module dominates. Seven carry 9 percent each, four carry 8 percent, and the opening context module is 5 percent. Click any module for what it covers.
The smallest module. Where AI security sits in an enterprise: using AI for security, securing AI systems and governing AI risk, how ownership splits across security, engineering, privacy, product and governance teams, and use cases such as enterprise chatbots and internal AI assistants.
The parts of an AI application (interfaces, prompts, models, APIs, retrieval systems and vector databases), trust boundaries and data flows, Zero Trust, XDR, EDR, SIEM and SOAR concepts, and AI threat modeling with OWASP LLM Top 10 mapping, MITRE ATLAS concepts and AI risk registers.
Python for log analysis, configuration review, evidence collection and control validation. Covers libraries such as Pandas, NumPy, Scikit-learn, Requests, BeautifulSoup, Scapy and OpenCV, security data formats like CSV, JSON, JSONL and YAML, regular expressions, and safe scripting patterns.
Securing prompt templates, orchestration layers, model APIs, sessions and guardrails. Secure design patterns such as context separation, least privilege, rate limiting, output validation, human approval workflows and audit logging, plus a secure AI SDLC and deployment readiness reviews.
Direct, indirect, multi-turn, role-play, encoded and instruction-conflict prompt injection, jailbreak methods, system prompt extraction and sensitive information disclosure. Then the defenses: output sanitization and validation, sandboxing, allowlists and user confirmation steps.
How retrieval-augmented generation moves data through chunking, embeddings, vector storage and retrieval, the risks that come with it (poisoned documents, retrieval manipulation, cross-user data leakage, vector data exposure), and controls like source allowlists, provenance tracking, access-aware retrieval and citation checks.
Protecting training, prompt, embedding and evaluation data from leakage, contamination and poisoning, verifying model artifacts and their sources, MLSecOps controls and secrets management in the ML pipeline, AI-based detection, model monitoring and adversarial machine learning risks.
Deployment patterns such as model endpoints, API gateways, containers, serverless workflows and vector databases. Service accounts, OAuth, workload identity, key rotation and tenant isolation, and abuse prevention through rate limits, quotas, cost controls, encryption, KMS and network isolation.
AI telemetry such as prompt logs, response logs, retrieval events and policy decisions, privacy-aware logging, SIEM integration, detection engineering for prompt injection and data exfiltration, and response steps including session reconstruction, containment, key rotation, document revocation and rollback.
NIST AI RMF, ISO/IEC 42001 and EU AI Act high-risk requirements, AI risk tiering and ownership models, privacy risk in prompts, outputs, logs, embeddings and fine-tuning data, and assurance artifacts such as risk registers, control matrices, model cards and an AI BOM.
AI red teaming methods, manual and automated adversarial testing, scope and rules of engagement, advanced threat vectors such as multimodal attacks, retrieval manipulation and excessive permissions, and writing technical and executive reports that tie findings to business impact.
An end-to-end AI security build in the course, with project themes such as a threat intelligence dashboard, SOC automation and LLM security monitoring. AI CERTs' certification scheme says capstone items on the exam are scenario-based multiple-choice questions, so there is no project to submit.
Modules and weights follow the AT-2102 exam blueprint on AI CERTs' AI+ Security Expert page and store listing, read October 21, 2026. AI CERTs' certification scheme document for AT-2102 (version 3.0, March 18, 2026) still lists an earlier ten-domain outline, so confirm the current outline with AI CERTs before scheduling. The course also has an optional AI Agents Security module that sits outside the blueprint.
The questions candidates ask most often when researching the AI CERTs AI+ Security Expert certification.
AI+ Security Expert is an intermediate AI security certification from AI CERTs, tested by exam AT-2102. It checks whether you can secure AI applications and use AI in security work, from threat modeling and prompt injection to RAG security, ML pipeline controls, secure deployment, monitoring, governance and red teaming. It is the middle credential in AI CERTs' three-tier security track, between AI+ Security Practitioner and AI+ Security Strategist.
Yes. The AI CERTs store lists AI+ Security Expert under exam code AT-2102 and states that it was formerly known as AI+ Security Level 2. AI CERTs' certification body site still publishes the certification under the older name with the same exam code. The sibling credentials were renamed the same way, to AI+ Security Practitioner (AT-2101) and AI+ Security Strategist (AT-2103).
The AT-2102 exam has 50 multiple-choice questions and a 90-minute time limit. You need 70 percent, which is 35 correct answers, to pass. You take it online through AI CERTs' own remote proctoring, and the 90 minutes include about five minutes to sign the candidate agreement and five minutes for the proctoring tutorial. You need a webcam and a quiet room with a single screen.
AI CERTs' current blueprint has 12 modules. AI Security Context, Scope and Opportunities is 5 percent. Architecture and Threat Modeling, Python Automation, GenAI Application Security Controls, Prompt Injection and LLM Risk Testing, RAG Security, AI Data and ML Pipeline Security, and Secure AI Deployment are 9 percent each. Monitoring and Incident Response, Governance and Compliance, Red Teaming, and the Capstone Project are 8 percent each.
No. AI CERTs' certification scheme sets no mandatory prerequisites for education, experience or prior certifications, and it says the level designation reflects the depth of the content, not a required progression. AI CERTs does recommend some grounding in computer systems, networking, cybersecurity basics, AI and machine learning concepts, and introductory Python, so AI+ Security Practitioner or similar study helps if you are new to the field.
As of October 2026, the AI CERTs store lists the AI+ Security Expert self-paced course at $495, which includes the online proctored exam with one free retake. That store page does not list a separate exam-only price. Training Camp's AI+ Security Expert boot camp includes the exam voucher in the course fee.
One year from the date AI CERTs issues the certificate. Renewal is by re-examination only. You pay the recertification fee and pass a recertification exam before the expiry date, and the credential is then valid for another year. If it lapses without an approved grace period, you have to go through the full certification process again.
AI CERTs says it operates as an independent certification body under ISO/IEC 17024:2012. Its accreditation page also says it is actively pursuing ANAB accreditation, so by AI CERTs' own account the credential is not ANAB-accredited yet. Check that page for an update before you rely on it to meet an employer requirement.
AI CERTs does not claim a DoD 8140 listing for AI+ Security Expert on its certification pages. If you need a credential for a DoD cyber position, check the current qualification matrix on the DoD Cyber Exchange and choose from the certifications listed there. See the current DoD 8140 work role paths.
AI CERTs publishes two outlines for AT-2102. Its certification scheme document, version 3.0 dated March 18, 2026, still lists an earlier ten-domain outline under exam code AIC-SEC-201, built around Python, machine learning, email and malware threat detection, network anomaly detection, AI-based authentication, GANs and AI-assisted penetration testing. The 12-module blueprint is the one AI CERTs shows on its current AI+ Security Expert page and store listing. Ask AI CERTs which outline applies before you book.
Both target hands-on AI security work, but the exams and upkeep differ. SecAI+ (CY0-001) has up to 60 questions in 60 minutes, including performance-based items, at a Pearson VUE center or online, and it lasts three years with continuing education. AI+ Security Expert is 50 multiple-choice questions in 90 minutes, online only, and you renew it by exam every year. Pick the one your employer or clients recognize.
AI CERTs links the certification to roles such as AI Security Analyst, Cybersecurity Engineer, SOC Analyst, Cloud Security Engineer, Penetration Tester and Security Consultant. It fits best for people already in one of those jobs whose work now includes securing LLM applications, RAG systems or ML pipelines.
Yes. AI CERTs requires a seven-day wait after a failed attempt before you schedule the next one. The self-paced bundle on the AI CERTs store includes one free retake, and further attempts go through AI CERTs' re-attempt process for a fee. Training Camp's boot camp also includes an exam retake.
Tell us if you're weighing AI+ Security Expert against SecAI+ or COASP, or planning AI security training for a team, and we'll help you map out the right path.