Safety and security for LLM systems
Prompt injection, jailbreaks, data leakage, over-powered agents, hallucination and fairness treated as engineering problems, with where each attack gets in, which controls actually cap the damage, and how to answer governance questions without reciting a statute.
What you'll be able to do
- Tell prompt injection from jailbreaking, and direct from indirect injection, by who attacks whom
- Apply the lethal trifecta test and least privilege so an agent stays safe when an injection succeeds
- Trace every path sensitive data can leak through an LLM system and put a control on each one
- Design guardrails and a red-teaming programme that account for latency, false positives and regressions
- Answer hallucination, fairness and governance questions with concrete controls and the right framework
Before this: agents-and-tool-use
Log in to read the rest of this lesson
Prompt injection, jailbreaks, data leakage, over-powered agents, hallucination and fairness treated as engineering problems, with where each attack gets in, which controls actually cap the damage, and how to answer governance questions without reciting a statute. This walkthrough runs about 18 minutes, with runnable code you can edit and re-run as you read. A new account gets 7 days of Pro free, no card required.
Log in to continueThe first module of every track is free to read on the web: see what's open in AI interviews.