Certified AI Safety & Security
Prompt injection basics
Module: Safety & Guardrails
Prompt injection is when untrusted text tries to override your instructions.
In practice: A web page saying 'ignore previous instructions' fed into a summary.
Try It Yourself
A web page saying 'ignore previous instructions' fed into a summary.
Lessons
▶️ 1. Prompt injection basics
▶️ 2. Defending against injection
▶️ 3. Avoiding data leakage
▶️ 4. Refusals and boundaries
▶️ 5. Reducing bias
▶️ 6. Content safety
▶️ 7. PII handling
▶️ 8. Responsible disclosure of limits
▶️ 9. Prompt injection
▶️ 10. Injection defences
▶️ 11. Jailbreak awareness
▶️ 12. Data exfiltration risks
▶️ 13. Output filtering
▶️ 14. Red-teaming prompts
▶️ 15. Least-privilege design
▶️ 16. Monitoring and logging
▶️ 17. Why AI ethics matter
▶️ 18. Bias and fairness
▶️ 19. Transparency
▶️ 20. Privacy and consent
▶️ 21. Misinformation risks
▶️ 22. Intellectual property
▶️ 23. Environmental and access issues
▶️ 24. Human accountability
▶️ 25. The Persona pattern
▶️ 26. The Template pattern
▶️ 27. The Recipe pattern
▶️ 28. The Flipped-Interaction pattern
▶️ 29. The Reflection pattern
▶️ 30. The Refinement pattern
▶️ 31. The Fact-Check pattern
▶️ 32. Combining patterns
▶️ 33. Prompts as code
▶️ 34. Prompt libraries
▶️ 35. Cost management
▶️ 36. Latency and performance
▶️ 37. Security and compliance
▶️ 38. Governance
▶️ 39. Monitoring quality at scale
▶️ 40. Change management
Course Home