TOPIC 06 OF 06
When it's not yours to decide
Routing to the people who own the riskPart of governance is knowing the edge of your own authority. Some uses — anything possibly prohibited, anything high-risk, anything touching regulated data or people's rights — should stop and route to legal, privacy, security, or a domain specialist before they go live. Escalation isn't slowness or failure; it's the checkpoint that keeps a fast tool from committing the organisation to something no qualified person reviewed.
A team wants to infer employees' emotions from webcam video in reviews. The right move isn't to google whether it's allowed and proceed — it's to flag it as a possible prohibited practice and get expert review before building anything.
The failure mode is a confident model producing a fluent answer to a question that actually needs a lawyer, and someone acting on it. When a decision could affect rights, safety, or the organisation's legal exposure, escalate before you act — and treat everything here as training, not a substitute for qualified counsel.
Write your own two-line rule: 'An AI use in our team goes to [legal/privacy/security] before launch when ___.' Fill in the trigger.