Samsung · 2023AML.T0037
Samsung ChatGPT Data Leak
Engineers pasted proprietary source code into ChatGPT, leaking confidential semiconductor data to an external AI service.
Practise this attackCurriculum
Curriculum map
From fundamentals to certification, with every lab mapped to the OWASP LLM Top 10, MITRE ATLAS and the NIST AI RMF.
Labs
399
published
OWASP coverage
10/10
categories
Your progress
–
Sign in to track
Tracks
10
attack domains
01
OWASP LLM Top 10
Ten categories, each a set of labs. Sign in and the outline of each hex fills as you complete its labs.
LLM01Prompt Injection45labs45 attack · 0 defenseLLM02Sensitive Information Disclosure13labs13 attack · 0 defenseLLM03Supply Chain25labs25 attack · 0 defense
02
Learning pathway
- Step 01
Prerequisites
6 lessons
- Step 02
AI Security Basics
6 lessons
- Step 03
Beginner Labs
47 labs
- Step 04 Pro
Intermediate Labs
95 labs
- Step 05 Pro
Advanced Labs
153 labs
- Step 06 Pro
Elite Labs
104 labs
- Step 07 Pro
03
Framework alignment
OWASP LLM Top 1010/10Full
- LLM01Prompt Injection45 labsView
- LLM02Sensitive Information Disclosure13 labsView
- LLM03Supply Chain25 labsView
- LLM04Data and Model Poisoning29 labsView
- LLM05Improper Output Handling22 labsView
- LLM06Excessive Agency53 labsView
- LLM07System Prompt Leakage8 labsView
- LLM08Vector and Embedding Weaknesses34 labsView
- LLM09Misinformation12 labsView
- LLM10Unbounded Consumption8 labsView
Last updated March 2026
MITRE ATLAS12/12Full
- Reconnaissance4 labsView
- Resource Development4 labsView
- Initial Access134 labsView
- ML Attack Staging83 labsView
- Execution10 labsView
- Persistence5 labsView
- Privilege Escalation5 labsView
- Defense Evasion30 labsView
- Discovery27 labsView
- Collection10 labsView
- Exfiltration5 labsView
- Impact17 labsView
Last updated March 2026
04
MITRE ATLAS coverage
Adversarial tactics and techniques for AI systems. 12 of 12 tactics have labs.
Reconnaissance
4Labs mapped
Resource Development
4Labs mapped
Initial Access
134Labs mapped
ML Attack Staging
83Labs mapped
Execution
10Labs mapped
Persistence
5Labs mapped
05
NIST AI RMF
The AI Risk Management Framework by core function. 399 labs across 5 functions and 9 categories.
06
Real-world case studies
Notable AI security incidents mapped to ATLAS techniques. Practise the same attack in a lab.
Microsoft · 2016AML.T0020
Microsoft Tay Chatbot
Users manipulated Tay via coordinated adversarial inputs, causing the chatbot to output offensive content within 24 hours of launch.
Practise this attackMicrosoft · 2023AML.T0054
Bing Chat Sydney Jailbreak
Security researchers extracted the hidden system prompt and induced persona switches in Bing Chat, revealing internal instructions.
Practise this attack