AI-system security
Adversarial inputs and vulnerable integrations can undermine an AI application’s intended behavior.
Reviewed · 2026-10-072 cited sources
Skip to Observatory contentPSYBORG / CONNECTED INTELLIGENCE
AI Risk Registry
Explore evidence-backed AI risk areas, human stakes, limitations and connected incidents, research and regulation.
5 records found · Page 1 of 1
Date filters use the record’s date or period, not its review date. Unknown dates are excluded by date filters. Related filters use direct, labeled relationships.
Biometric misidentification
An incorrect identity match can become a harmful human decision.
Reviewed · 2026-10-072 cited sources
Privacy in AI deployments
Collecting and using personal data creates questions about purpose, access and retention.
Reviewed · 2026-10-072 cited sources
Synthetic-media impersonation
Generated voices and images can borrow someone else’s identity to mislead an audience.
Reviewed · 2026-10-072 cited sources
Unreliable generated information
Fluent answers can contain errors that readers fail to notice.
Reviewed · 2026-10-072 cited sources