My PAPA (Practical AI Pentest Associate) Review

Semi-recently, TCM Security released their own AI penetration testing certification — the Practical AI Pentest Associate (PAPA). With the prevalence of AI rapidly increasing, I feel an accreditation in this field would be great to have. In this blog, I will recount my experience with the course and exam, which were both great in my opinion. I'm not exactly sure what I can — and more importantly can't — say about the exam. I wouldn't want a trenchant email from TCM, saying they're revoking my certification and pursuing legal action. However, I'll try to articulate my personal experience along with any advice I have for people attempting this cert. I particularly enjoyed TCM's style of exam; I believe they pride themselves on their realistic approach to assessing students. The exam is not CTF-like, but rather comparative to a real-world engagement.

The Course

The certification comes with two courses — AI Fundamentals 100 and AI Hacking 101. They are both instructed by Andrew Bellini, whom I thought did an incredible job at teaching the content. These two courses include roughly 10 hours worth of material.

The labs were fun and engaging, but there was one caveat — they were all self-hosted. This typically wouldn't be a problem; however, my old laptop did not appreciate the volume of messages I was sending to Ollama.

The AI fundamentals course covered LLMs, natural language processing, neural networks, and self-hosting. Overall, it was comprehensive and engaging. Additionally, TCM Security provides access to this course for free. For a free course, the content is genuinely amazing.

For me, AI Hacking 101 is where things got interesting. Its syllabus includes:

• Prompt Injection
• Jailbreaking
• Reconnaissance
• Excessive Agency
• Retrieval Augmented Generation (RAG)
• Harmful & Wasteful Output
• Role-Based Access Control

Preparation for the Exam

I first began the coursework in December 2025. I figured I should finish the courses, practice for a while, then sit the exam while the content was still fresh in my mind. Instead, due to my complete lack of forward thinking, it took me until July 2026 to take the actual exam. In that period of time, I got the opportunity to participate in the TCM and TryHackMe AI CTFs, which were good preparation. TryHackMe now has an increasing amount of AI-related rooms.

The Exam

The exam lasts four days. You have two days to complete the penetration test and a further two days to write a professional report. You need at least 75% to pass. Since the exam is not CTF-like, I was constantly stressing about whether I had done enough to secure 75%. Honestly, I probably scraped by with a 76%. While sitting the exam, I didn't know how many vulnerabilities were present. In the end, I methodically tested for each vulnerability covered in the course and prayed it was enough. One thing about TCM's exams is that you don't have to sit them under strict, proctored conditions. At one point in time, I hit a wall, so I sat on the floor of my room and watched YouTube shorts. While I don't think this exam was too intensive, I found taking breaks to clear my head was important.

After completing the penetration test, I only took one day to write my report. Writing professionally is certainly not my strong-suit, so this was stressful for me. When you begin the exam, you are given a RoE (rules of engagement). Here, there was a template for the final report. My report had the following structure:

• Executive Summary
• Test Details (e.g. testing period, scope, tools used)
• Technical Findings
• Final Notes / Conclusion

For each technical finding, I wrote what it was, its severity, remediatory action, evidence of vulnerability, and steps to reproduce it.

Finally, I submitted my report. After a week of anticipation, I passed.

Even though exams are inherently stressful, I found it fun. Yes, the process had its low ebbs, but even if I failed, I would have leant a lot and enjoyed taking the exam. If you're interested in cybersecurity, I definitely recommend this certification.