What should the rules and constraints of an AI component define?
Why C is the answer
What it may do, and what it must refuse
Solution rules define limits, such as allowed actions and when to escalate.
In detail
- Rules define what the component may do, and what it must refuse.
- State refusals and escalation paths in writing.
Use case
An illustrative scenario, not a real incident.
An agent may issue refunds up to a set amount, and it must hand larger ones to a person.
See it in different ways
mindmap
root((Agent types))
Prompt and response
User asks in chat
Task
User starts it
Autonomous
Event starts it
No user present
Multi-agent
Specialistsflowchart TD
Q{"Is the action allowed?"}
Q -->|"Within limits"| A0["Proceed"]
Q -->|"Beyond limits"| A1["Escalate"]
Q -->|"Forbidden"| A2["Refuse"]Key point: Rules define limits.
Why the other options are wrong
- A. The name of its project sponsor
- A sponsor's name defines no behavior.
- B. The order of the columns in its log file, and the file name it writes to
- Log formats define no behavior.
- D. The model version that it runs on
- A model version is a technical setting.