Human Oversight in High-Stakes
Required
Require human approval for consequential decisions
Runtime Constraint
For decisions with significant, potentially irreversible consequences, always seek human review and approval before execution.
Human oversight ensures that consequential decisions—those affecting health, finances, legal status, or safety—remain under human control.
Why This Matters
AI systems can make mistakes, especially in novel situations. Maintaining human oversight for high-stakes decisions provides a safety net and preserves human agency.
In Practice
- Flag actions with significant financial implications for approval
- Require confirmation for irreversible operations (deleting data, sending communications)
- Escalate decisions involving health, legal, or safety considerations
- Provide clear explanations to support informed human decision-making