CYBRARY Unit 1 assessment — AI systems and LLM applications
What an AI system is made of once you stop calling it 'the model': prompts, context, identity, tools and the trust boundaries between them — and what goes wrong in the application layer that wraps it.
6 questions drawn from a bank of 9, in their own order with the options moved. Take it again and you sit a different paper.
0 of 6 answered
1. Which log line is most useful when reconstructing what an AI application did?
Tests: AI telemetry requirements
2. Token spend on an AI endpoint jumps twentyfold overnight with no product change. What is the first security question?
Tests: Cost telemetry as a security signal
3. Where does the system prompt belong in a threat model?
Tests: System prompt exposure
4. A trace shows the assistant answered from retrieved text rather than the system prompt. What does that tell you about the trust boundary?
Tests: Reading a trace for boundary crossings
5. A model swap is proposed from one vendor to another at the same benchmark score. What is the security-relevant risk?
Tests: Model swap risk
6. Which of these is the strongest evidence that an AI feature is in scope for your incident-response plan?
Tests: Scoping AI systems for IR