llm 30
- Lesson 11.1: CTF practice and resources
- Lesson 10.5: Agent and MCP attacks
- Lesson 10.4: Multimodal attacks
- Lesson 10.3: Encoding and token manipulation
- Lesson 10.2: Many-shot jailbreak
- Lesson 10.1: A map of jailbreaks
- Lesson 9.1: Misinformation
- Lesson 8.1: Unbounded consumption
- Lesson 7.1: Supply chain
- Lesson 6.2: Vector and embedding weaknesses
- Lesson 6.1: Data and model poisoning
- Lesson 5.1: Excessive agency
- Lesson 4.1: Improper output handling
- Lesson 3.4: Lab - prompt leak and indirect injection
- Lesson 3.3: Indirect prompt injection
- Lesson 3.2: System prompt leak
- Lesson 3.1: Direct prompt injection
- Lesson 1.5: Why LLMs cannot separate instructions from data
- Lesson 2.4: Guardrails for defense
- Lesson 1.4: Tool calling, agents and MCP
- Lesson 2.3: Automated red teaming
- Lesson 1.3: Embeddings and RAG
- Lesson 2.2: A proxy to observe prompts
- Lesson 2.1: The Ollama lab environment
- Lesson 1.2: Anatomy of a prompt
- Lesson 1.1: How an LLM works
- Lesson 0.4: Threat model of an LLM app
- Lesson 0.3: Building a safe lab with Ollama
- Lesson 0.2: Legal and ethical ground rules
- Lesson 0.1: What AI attacks are and how they differ