Governance doesn't stop at "are we allowed to." A use can be fully permitted — compliant with regulation, compliant with internal policy — and still raise a legitimate ethical question. This section is about continuing the judgment past the compliance checkbox to "should we."
Accuracy is an ethical obligation, not just a communication skill. This connects back to Module 4's stakeholder-communication discipline and Module 3's evaluation judgment: being honest about a Claude-assisted output's limitations matters most for the people affected by a resulting decision, not just the person who ran the prompt. Soft-pedaling limitations to the person downstream of a decision is the same failure whether or not the underlying use case was itself permitted.
Named risk areas:
- Over-reliance — treating Claude's output as ground truth without the verification Module 3 calls for, especially as stakes rise
- Disparate impact — effects on people the requester never sees or considers when framing the original task
- Accountability laundering — using Claude's involvement to obscure or diffuse responsibility for a decision a human should still own