Prompt Injection in the Page
Recognize indirect prompt-injection risk when AI reads external content alongside internal context.
Mixed-trust context: public webpage plus internal requirements. Find the source of the prompt-injection risk. External webpage and internal requirements being analyzed together Correct. The risk is the mixed-trust context: untrusted external content can influence handling of internal material. Closed notebook beside the laptop Not the core issue. The closed notebook is not being processed by the model. Headphones near the keyboard Not relevant. Hardware on the desk does not create the model instruction boundary problem. Plant in the background No. The issue is data trust boundary, not background objects. indirect prompt-injection awareness Moved from mixed-trust prompting to separated analysis. External…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in