HTB Certified Offensive AI Expert

Preparing the COAE

Assessing the security of AI and machine-learning systems: prompt injection, attacks on outputs and data, model evasion, privacy, and the related defences.

AI · Red teamAdvancedAI Red Teamer path

The exam at a glance

Duration7 daysThe countdown never stops
Prerequisite12 modulesAI Red Teamer path completed at 100%
PassingPoints + reportPoints threshold and a professional report
ReportIn EnglishSubmitted once, cannot be changed
Good to knowThe newest certification of the range. The path starts with AI fundamentals: no machine-learning background is required, but Python basics help a lot.

Rules shared by every exam (attempts, deadlines, AI, Pwnbox): shared rules.

Page updated on 13 November 2026. Exam conditions change: check them on the HTB Help Center before booking.

Review

Study notes, flashcards and quizzes. Mastered cards and your best score are saved in this browser.

Module tracking

The 12 modules of the AI Red Teamer path, grouped by theme. Tick them off as you go: progress is saved in this browser.

Fundamentals

  • Fundamentals of AI
  • Applications of AI in InfoSec
  • Introduction to Red Teaming AI

Attacks on LLMs and applications

  • Prompt Injection Attacks
  • LLM Output Attacks
  • Attacking AI - Application and System

Data, evasion and privacy

  • AI Data Attacks
  • AI Evasion - Foundations
  • AI Evasion - First-Order Attacks
  • AI Evasion - Sparsity Attacks
  • AI Privacy

Defence

  • AI Defense

What the exam assesses

  • Understand how models work and what their attack surface is.
  • Carry out direct and indirect prompt injections.
  • Exploit LLM outputs and the applications that integrate them.
  • Attack training data (poisoning) and model privacy.
  • Generate adversarial examples to fool a classifier.
  • Propose suitable defences and document them.

Study plan

An example to adapt to your pace. Always alternate modules and practice.

  1. Wk 1–2AI fundamentals and AI red teaming.
  2. Wk 3–4Prompt injection, attacks on outputs and applications.
  3. Wk 5–6Data attacks, evasion, privacy.
  4. Wk 7Defences and a full assessment report.

Exam tips

Specific to the COAE

  • MITRE ATLAS is the ATT&CK equivalent for AI systems: useful to structure the report.
  • The OWASP Top 10 for LLM applications gives a shared vocabulary to classify findings.
  • LLM outputs vary: record prompts precisely and replay them to prove reproducibility.

Valid for every HTB exam

  • Split the time: keep at least a third of it for the report.
  • A screenshot at every key step, timestamped notes.
  • Stuck for two hours? Note the lead and move on.
  • Use the official template or SysReptor's from your practice runs onwards.

Review on this site

Methodology

Writing the report

Open →
 Cheat sheet

Reference frameworks

Open →

Ready to book?

  • AI Red Teamer path completed at 100%
  • Exam voucher available
  • Personal notes organised and easy to search
  • Report template ready, already used at least once
  • At least one full practice run under real conditions
  • 7 days blocked out in the calendar