Module 6 Summary: Control the Risk or Stop
Consolidate the module decisions and keep the reusable resource available for later work.
The responsible-use sequence
- Identify affected people and possible harm.
- Define the purpose and minimise information.
- Treat external content as untrusted.
- Limit permissions, tools, and actions.
- Check bias, fairness, accessibility, and challenge routes.
- Resolve or escalate copyright, licence, attribution, disclosure, and ownership questions.
- Assign named owners and meaningful human decisions.
- Record controls, escalation, and an explicit stop condition.
Main lessons
- Task usefulness does not grant data permission.
- Removing one name may not prevent identification.
- Prompt injection risk is managed through system design and limited impact, not a magic instruction.
- More agent access means more possible harm.
- Average performance can hide exclusion.
- Accurate content can still be inaccessible.
- Citation is not automatically permission.
- A generic human in the loop is not meaningful oversight.
- Not using AI may be the right result.
Important review status
This module remains subject to specialist legal, privacy, copyright, and security review before public release. The knowledge check tests the cautious course method, not legal conclusions.
What comes next
Module 7 combines task choice, specification, verification, safeguards, human decisions, testing, and reflection in one capstone workflow.
Keep this rule:
If you cannot name the affected people, permitted data, responsible owner, and stop condition, do not proceed.