AAISM: Advanced in AI Security Management

Question

An organization uses an AI model to optimize traffic signals and forecast congestion. Because of misconfiguration and insufficient validation, the system excessively prioritizes low-traffic areas, resulting in gridlock in major cities. What should be done FIRST to address this problem?

Explanation

An AI system producing outcomes inconsistent with its intended use and causing active operational harm should be promptly disengaged or deactivated to contain the impact. Remediation such as validation, rollback assessment, or retraining can occur after the unsafe behavior has been stopped. Transparent communication is also an appropriate component of incident response. NIST’s AI Risk Management Framework Playbook identifies mechanisms to supersede, disengage, or deactivate AI systems whose performance or outcomes are inconsistent with intended use.

Learn more

NIST AI Risk Management Framework Playbook — Manage

AAISM: Advanced in AI Security Management

AstroTutor

Notes

Readiness

QuestionQ62

Community Discussion