Waymo Robotaxis Allegedly Contributed to Traffic Gridlock During San Francisco PG&E Power Outage

Waymo's autonomous vehicle fleet reportedly contributed to traffic congestion in San Francisco after a PG&E substation fire cut power to nearly one-third of the city on December 20, 2025. As traffic signals went dark, Waymo vehicles requested remote confirmation checks at scale, creating delays that added to gridlock and prompted city officials to order the fleet off the road. Waymo later paused service and announced software and emergency-response updates.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The incident represents a lack of robustness under extreme, large-scale infrastructure failure conditions, where the AI system's safety protocols became overwhelmed and caused the vehicles to stall.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The traffic congestion was caused by the Waymo Driver AI system's safety protocols triggering excessive remote confirmation requests during a post-deployment power outage, which was an unintentional outcome of its design.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI system is used in critical infrastructure (specifically road traffic management and transport), which has significant implications for public safety and emergency responder access during a crisis.

AI system and alleged parties

• AI system: Waymo (Google)
• AI purpose: Autonomous Driving; Navigation Assistant
• Behaviour type: Autonomous
• Alleged developer: Waymo
• Alleged deployer: Waymo
• Alleged harmed parties: Waymo passengers, San Francisco road users, San Francisco emergency services, San Francisco drivers, General public of San Francisco, General public, Emergency responders

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Minor, indirect Minor
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Infrastructure

Reported: The report describes city-wide traffic gridlock and blocked intersections, but does not explicitly describe physical damage to infrastructure caused by the AI system.

People affected

• Occurrences reported: 1
• People reportedly exposed: 130000

Potential causes

Management

• Insufficient Risk Assessment: Management failed to anticipate the impact of a large-scale power outage.
• Underprepared Emergency Protocols: Emergency plans did not account for coordinated fleet-wide pull-overs.

Technology

• Remote Confirmation Backlog: High volume of requests overwhelmed remote operators, delaying vehicle response.
• Lack of Local Outage Context: Vehicles lacked regional power outage context to navigate autonomously.
• Conservative Fallback Design: Vehicles default to stopping and requesting checks at dark signals.

Data Inputs

• Missing Real-Time Outage Data: Vehicles had no direct data feed about regional grid power status.
• Dark Traffic Signal Detection: Encountering dark signals triggered safety checks instead of decisive driving.

Human Factors

• Operator Overload: Remote assistance staff could not handle the sudden spike in requests.
• Manual Traffic Control Conflicts: Vehicles struggled to coordinate with manual police traffic management.

Process and Methods

• Inadequate Scaling Protocols: Confirmation checks worked for small outages but failed at city-wide scale.
• Delayed Service Suspension: Service was suspended only after gridlock and mayoral intervention occurred.

Regulatory Environment

• Lack of Emergency AV Guidelines: No established regulatory standards for AV behavior during power outages.

Information quality

• Classification confidence: High
• Reason for confidence: The reports from multiple reputable news sources and Waymo's official blog provide consistent, detailed accounts of the technical mechanism behind the failure (the remote confirmation check backlog) and the operational impact (traffic congestion and service suspension). There is high agreement on the facts of the incident.
• Ambiguities identified: None of significance; the technical explanation of the backlog and the timeline of the service suspension are clearly laid out.
• Alternative interpretations: One could argue the primary cause was the utility power outage rather than the AI, but the reports clearly focus on how the AI's specific safety protocols uniquely contributed to the traffic gridlock.

During a San Francisco power outage, Waymo's autonomous vehicle fleet suffered a systemic safety-protocol failure, stalling at dark intersections and causing gridlock that impeded emergency responders. While local in scope, the incident highlights minor but notable vulnerabilities in critical transportation infrastructure when integrating autonomous systems that rely on remote human oversight.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The vulnerability of autonomous fleets to infrastructure failures represents an ongoing strategic and regulatory concern rather than an active crisis.
• Autonomy: Human-supervised. The vehicles operate autonomously but rely on remote human confirmation for complex scenarios, creating a bottleneck when the system was overwhelmed.
• Novelty: Evolved capability. While autonomous vehicle stalls are established events, the scale of failure triggered by a widespread power outage represents an evolved systemic vulnerability.

Impact by dimension

• Physical security: Minor. The stalling of autonomous vehicles during a power outage caused localized gridlock and temporarily impeded emergency responders, representing a minor impact on critical transportation infrastructure.
• Information security: Negligible. No information warfare, intelligence compromise, or systematic information manipulation occurred during this incident.
• Sovereignty: Minor. Local government functions, specifically emergency response routing, were temporarily disrupted, but the issue was quickly managed by city officials ordering the fleet off the streets.
• Economic security: Negligible. The economic impact was limited to temporary suspension of Waymo's local ride-hailing services and minor commuter delays, with no strategic technological threat.
• Societal stability: Negligible. The incident did not threaten social cohesion, civil liberties, or population safety on a national scale.