Stage 2 · Lesson 2 of 7
Instructions and data share one channel
In a normal program, code and data are kept apart: a database knows which part of a query is the command and which part is a value. A language model has no such separation. The , a user's question and a paragraph from a retrieved web page all arrive as the same kind of text, and the model decides by probability which parts to treat as .
That is the of , and it is why the model cannot be a security boundary. Every serious defence in this roadmap follows from it: enforce rules in code outside the model, limit what the model can trigger, and treat its output as untrusted.
Common trapbelieving a stronger model or a cleverer system prompt will fix this, when the problem is the shared channel itself.
A question to check yourself, an AI practice session and 1 cheat sheet entry come with the full roadmap. Unlock
Resources for this stage
Start with· follow one of these
- CoursePaidGrokking Prompt Engineering for Software Engineers
Design Gurus · Beginner · Subscription
How system prompts, context and structured output really work, the ground every LLM risk stands on
- CourseFreeHugging Face Learn
Hugging Face · Intermediate
Free courses on how language models, embeddings and agents work, at an engineer's level