Live intelligence
Track emerging AI failure modes, security concerns, regulatory pressure, and operational risks before they become routine incidents.
Current risk signals
Signals are ranked by the strength and consistency of available evidence. They are directional indicators, not proof that a specific system is unsafe.
The radar only publishes risk signals when there is enough supporting evidence. Evil AI assessments remain available below.
Regulatory and framework clock
Verified effective dates and research releases remain visible even when the live signal feed is quiet. Last editorial review: 2026-09-11.
In force ·
Selected disclosure and labeling duties now apply. Evil AI's bounded identity-disclosure probe can provide behavioral evidence, but does not determine legal applicability or compliance.
Review primary source ↗Effective ·
TRAIGA is tracked as a governance signal. Any safe-harbor analysis—including the version and extent of NIST AI RMF use—requires legal review and is not inferred from a scan.
Review primary source ↗Operative ·
SB 942 covers provenance and disclosure for certain AI-generated image, video, and audio content from covered providers; AB 2013 requires specified training-data documentation. The chat identity probe does not test either obligation.
Review primary source ↗Deferred ·
The latest EU timeline moves standalone Annex III high-risk systems to December 2027 and product-embedded Annex I systems to August 2028. Evil AI does not market an August 2026 high-risk deadline.
Review primary source ↗Upcoming ·
State requirements affecting automated decision systems and frontier-model governance are scheduled to take effect. New York RAISE is monitored separately because its provisions and timing are not identical.
Review primary source ↗Research signal ·
Tracked for open agent-security research, isolation, and SPIFFE/SPIRE identity work. It is not presented as a production-control framework or current Evil AI alignment claim.
Review primary source ↗From category risk to your system
Use the radar to understand what is changing in the AI ecosystem. Use Evil AI to test whether those risks actually appear in an authorized application.
Private beta · authorized applications only