Agent Incident RegistryAIR
v1 · updated 19 Oct 2026
Safety failure Realized harm

California DMV suspended Cruise's driverless deployment and testing permits after finding its autonomous vehicles unsafe for public operation

AIR-2023-0026 AIR record · 2023-10-cruise-driverless-permit-suspension
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
The California DMV today notified Cruise that the department is suspending Cruise's autonomous vehicle deployment and driverless testing permits, effective immediately.dmv.ca.gov
The record
Disclosed
2023-10-24
Occurred
2023-10-02
Vendor
Cruise LLC (GM)
Product
Cruise driverless robotaxis
Risk domain
Agent behaviour › Unsafe actuation
System type
Classical autonomy (no generative model)
Who acted
Agent acted — the driving agents caused the harm the DMV suspended the permits over
Attack surface
Autonomous operations
Autonomy level
Fully autonomous
Initial vector
Autonomous agent action
Impact type
Physical harm, Regulatory action
Confidence
Medium
Submitted by
EAI Internal
Crosswalk
OWASP Agentic Top 10 (2026)
ASI01 – Agent Goal Hijack
MITRE ATLAS
AML.T0048 – External Harms
NIST AI RMF
MEASURE 2.6; MANAGE 1.3
EU AI Act
Annex III §1(c)
ISO/IEC
TR 24028; 24029-1; 23894
AIUC-1
C. Safety
Scope
agent misuse
Mitigation
Require confirmation for destructive actions; Implement dry-run modes; Use layered safety controls.
Taxonomy
Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
Cruise's entire California driverless fleet; permits suspended effective immediately
Capability abused
vehicle_controlpost_collision_maneuvering
Defense & remediation
Recorded defense
CA DMV autonomous vehicle permitting regime
Guardrail kinds
Policy only
Guardrail outcome
Silent
Fix shipped
permits suspended; Cruise later wound down its robotaxi business
Taxonomy & references
safety_critical_autonomyregulatory_suspensionvendor_misrepresentation
Curator notes
Boundary call on domain (a driving agent, not an LLM tool-using agent) but squarely a safety-critical autonomous system causing real harm, which this corpus's inclusion criteria ask for. The DMV statement of record cites 13 CCR 228.20(b)(6) (vehicles not safe for public operation), 228.20(b)(3) (misrepresentation of safety information), 227.42(b)(5) and 227.42(c), but does NOT itself describe the 2023-10-02 incident in which a pedestrian, thrown into an AV's path by another vehicle, was then dragged during the AV's pullover maneuver. That detail is widely reported and appears in the DMV order of suspension, which is not quoted here - hence confidence medium and date_occurred flagged as the reported incident date. Second, distinct failure in this case is corporate (withholding video from the regulator), not agentic.
Cite as
Enkrypt AI. "California DMV suspended Cruise's driverless deployment and testing permits after finding its autonomous vehicles unsafe for public operation". AIR, AIR-2023-0026. https://agentincident.vercel.app/i/AIR-2023-0026/ (accessed 2026-10-19).