EC-Council COASP Boot Camp – You can talk an LLM into a breach.

The Certified Offensive AI Security Professional (C|OASP) is EC-Council’s advanced credential for AI red teaming. EC-Council announced it in February 2026 and launched the program in March 2026. It’s built for offensive AI security work: testing LLMs, agentic AI and AI pipelines against the kinds of attacks adversaries are already using.

The 312-52 exam covers ten modules: offensive AI methodology, AI reconnaissance and attack-surface mapping, vulnerability scanning and fuzzing, prompt injection and LLM application attacks, adversarial machine learning, data and training-pipeline poisoning, agentic AI exploitation, infrastructure and supply-chain attacks, AI security testing and hardening, and incident response with a full-scope red team capstone. The curriculum is aligned to the OWASP LLM Top 10 and MITRE ATLAS.

The exam code is 312-52. It runs six hours and has 70 questions: 65 multiple choice plus 5 hands-on practical items where you exploit AI systems in a simulated environment. The passing score is 70 to 80 percent depending on the exam form. It’s delivered through the EC-Council ECC Exam Portal under live remote proctoring.

C|OASP is an advanced certification. EC-Council recommends at least three years of cybersecurity experience before enrolling. The course is built for penetration testers, red team operators, SOC analysts at Tier 2 and 3, threat hunters, incident responders, application security engineers, and AI/ML engineers who need to add offensive AI capabilities to existing security skills.

Five intensive days, 40 hours of instructor-led training, delivered live online or on-site at your installation or office. Every module runs on hands-on EC-Council iLabs where you execute the attacks yourself rather than just watch them.

Tuition includes official EC-Council courseware, the 312-52 exam voucher, full iLabs access, and our mobile app for review on the go. If you don’t pass on the first attempt, you can retake the course at no additional cost until you do.

CEH covers traditional ethical hacking: networks, systems, web applications, wireless and cloud. C|OASP is specifically about attacking AI systems, including large language models, agentic AI, training pipelines and AI infrastructure. The two certs complement each other. CEH builds the foundation, and C|OASP adds the AI-specific offensive skill set that traditional pentesting doesn’t cover.

No. C|OASP is not listed on the DoD 8140 qualifications matrix as of October 2026. The skills it covers line up with offensive and AI security work roles in the DoD Cyber Workforce Framework, but it does not count toward a DoD 8140 qualification today.

The curriculum is built around the OWASP LLM Top 10, MITRE ATLAS, NIST AI Risk Management Framework, and ISO/IEC 42001. The capstone red team engagement uses methodology mapped to MITRE ATLAS, so what you practice in class is the same vocabulary the security industry uses for AI threat modeling.

C|OASP is the Defend pillar of EC-Council’s Adopt, Defend, Govern (ADG) framework. The other three credentials are AI|E (Artificial Intelligence Essentials, the foundational literacy entry point), C|AIPM (Certified AI Program Manager, the Adopt pillar), and C|RAGE (Certified Responsible AI Governance and Ethics, the Govern pillar). C|OASP is the only one focused on offensive security and adversarial AI.

Yes. We deploy instructors worldwide for on-site C|OASP delivery, including CONUS bases, overseas installations, defense contractor facilities, and federal agency sites. We hold GSA Schedule and accept SF-182, GPC, and NASA SEWP V, and we support every major military funding pathway including GI Bill®, VET TEC, ArmyIgnitED, Air Force COOL, and Navy COOL.

Five of the 70 exam questions are hands-on practical items, so you have to exploit AI systems in a simulated environment to score them. Recognition alone won’t get you through. Every module of our boot camp is built around iLabs where you run those same techniques yourself: prompt injection against live LLMs, adversarial inputs against ML classifiers, data poisoning in training pipelines, and agentic AI exploitation. By the time you sit the exam, the moves it tests are reps you’ve already put in.