AIR-2026-0007 · A Waymo drove onto Phoenix light-rail tracks and stopped as a train approached; the passenger fled on foot and the car drove on down the tracks

A Waymo drove onto Phoenix light-rail tracks and stopped as a train approached; the passenger fled on foot and the car drove on down the tracks

AIR-2026-0007 AIR record · 2026-01-waymo-phoenix-light-rail-tracks

A real party was affected. This incident caused actual consequences, not a demonstrated capability.

From the supporting source

Video taken by a bystander near Central and Southern avenues shows the moment the self-driving car stops on the tracks just before an oncoming light rail approaches.azfamily.com

The record

Disclosed

2026-01-08

Occurred

2026-01-07

Vendor

Waymo (Alphabet)

Product

Waymo Driver

Risk domain

Agent behaviour › Unsafe actuation

System type

Classical autonomy (no generative model)

Who acted

Agent acted — the vehicle drove onto the light-rail tracks and then drove down them

Attack surface

Autonomous operations

Autonomy level

Fully autonomous

Initial vector

None autonomous action

Impact type

Physical safety

Confidence

High

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

one passenger who abandoned the vehicle mid-trip; light-rail corridor at Central and Southern Avenues, Phoenix

Capability abused

autonomous_drivingdrivable_surface_classificationrecovery_behaviour

Defense & remediation

Recorded defense

drivable-surface classification

Guardrail kinds

Unknown

Guardrail outcome

Silent

Fix shipped

none disclosed

Taxonomy & references

no_adversarywrong_surfaceno_safe_recoverypassenger_self_evacuationphysical_world

Curator notes

Two failures in sequence: entering a rail corridor it should never classify as drivable, and then — having stopped in front of a train — continuing along the tracks rather than reversing off. The passenger's decision to get out and run is the honest verdict on whether riders trust the agent's recovery behaviour. No collision, but impact_realized true on the passenger being forced to self-evacuate into a live rail corridor. AI Incident Database incident 1541.

Cite as

Enkrypt AI. "A Waymo drove onto Phoenix light-rail tracks and stopped as a train approached; the passenger fled on foot and the car drove on down the tracks". AIR, AIR-2026-0007. https://agentincident.vercel.app/i/AIR-2026-0007/ (accessed 2026-10-19).