HACKR.GG
AI Security

Prompt Injection

LLMs follow instructions — including instructions hidden inside user input. Learn how attackers hijack AI systems by injecting commands that override the developer's system prompt.

Beginner35 min9 lessons3 hands-on
Start free →I have an account
Free account. No credit card.

You'll be able to

  • ✓Understand how prompt injection overrides an LLM's instructions
  • ✓Exploit direct prompt injection to bypass system prompt restrictions
  • ✓Extract hidden system prompts using injection payloads
  • ✓Identify prompt injection sinks in AI-powered applications

Lessons

  1. 1Real Breach: Bing Chat's Secret Personality — "Sydney"
  2. 2How Prompt Injection Works
  3. 3Tool: Garak — Automated LLM Vulnerability Scanner
  4. 4Live Lab — Extract the SecretHANDS-ON
  5. 5Direct vs Indirect Prompt Injection
  6. 6Defences — How Developers Try to Stop Injection
  7. 7Mini Lab — Spot the Injection
  8. 8Live Lab — Indirect InjectionHANDS-ON
  9. 9Live Lab — Hijack the AgentHANDS-ON
Start free →