Module 1 Knowledge Check (Self‑Check)
This self‑check is designed to confirm you can identify the most common AI risk patterns before you move on.
Instructions
- Answer each question. Then read the explanation.
- If you miss two or more, re-read the lesson and revise your “risk intake” checklist.
Questions
-
Which AI failure mode is most associated with made‑up cases, statutes, or quotes?
- A. Bias
- B. Hallucinations
- C. Encryption failure
- D. Conflicts of interest
Suggested answer: B. Hallucinations.
Why: Generative models can produce plausible text that is not grounded in real sources. -
True or False: If an answer “sounds right,” it is safe to rely on it for court filing content.
- True
- False
Suggested answer: False.
Why: Court-facing work requires primary-source verification (citations, quotes, facts, and procedural posture). -
Which factor most increases confidentiality risk when using AI tools?
- A. The tool is used only for formatting
- B. The input includes client identifiers or privileged strategy
- C. The prompt is short
- D. The output is a bulleted list
Suggested answer: B.
Why: Sensitive data + third-party processing increases the risk of unauthorized disclosure or retention. -
Which control best addresses “automation bias”?
- A. Longer prompts
- B. Mandatory human verification and sign‑off for high‑impact outputs
- C. Using AI for everything
- D. Turning off grammar checking
Suggested answer: B.
Why: The control is procedural: require independent verification and accountability. -
Which is the safest default rule for staff?
- A. Send AI output directly to clients if it is labeled “draft”
- B. Use AI for legal advice if the client asks for speed
- C. Use AI only within approved tools and always route high‑risk outputs for attorney review
- D. Let AI decide what is privileged
Suggested answer: C.
Why: Role boundaries and supervision are core protections in legal practice.
If you answered fewer than 4 correctly: revisit the risk matrix lesson and write down your firm’s “high-risk” list (court filings, advice, privilege, deadlines, sensitive data).