Description
Policy defining permitted and prohibited uses of AI systems, including boundaries for autonomous decision-making and human involvement.
Implementation Guidance
Permitted and Prohibited Uses
Define explicit categories: approved use cases (content drafting, code assistance, data analysis), restricted use cases requiring approval (customer-facing decisions, HR screening), and prohibited uses (social scoring, unauthorized surveillance, weapons targeting).
Decision Boundaries
Specify thresholds for autonomous vs. human-assisted AI decisions based on impact level. High-impact decisions (credit, employment, healthcare) require human-in-the-loop. Document escalation paths when AI outputs fall outside acceptable confidence ranges.
Employee Guidance
Provide practical examples for each user role: what engineers can use AI for, what analysts should avoid, how managers should review AI-assisted recommendations. Include data handling rules for AI tool inputs.
Evidence Requirements
- Acceptable use policy with use case classifications
- Decision boundary matrix (autonomous vs. human-in-the-loop)
- Employee acknowledgment records
- Use case approval requests and decisions