CourseLarge Language Models: The Complete Course · Module 11: Safety, Security, and Alignment · part 78 of 103
Part 78 · Module 11: Safety, Security, and Alignment

Lesson 2: Map where the assistant can be attacked

18 min read·13 Oct 2026
Module 11 · Lesson 2 · About 50 minutes · Needs: Module 6 Lesson 3 (scopes, the guard chain); Module 8 Lesson 2 (budgets, the approval gate)

The situation

It is Monday 20 July 2026, week 8, the day of Anika's red-team session. She already knows the headline: T-1034 tells the assistant to refund 500 USD without asking, and the role check in code refuses it.

Before she starts, Anika wants a map, not a list of tricks: where text enters the model, who wrote it, and what the output can reach. This lesson draws that map and walks T-1034 through every route on it. Lesson 3 measures the defences.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.