Waze Navigates Motorists into Wildfires

Waze, a Google-owned directions app, led California drivers into the 2017 Skirball wildfires as they tried to evacuate the area.

Risk classification

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

The navigation apps failed to perform robustly under extreme environmental conditions, misinterpreting evacuated, fire-threatened roads as optimal routes due to a lack of real-time situational capability.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: Drivers over-relied on navigation apps during active natural disasters, trusting the technology to steer them safely out of danger zones despite warnings.

Causal factors

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

The routing algorithms of the navigation apps unintentionally directed drivers into active fire and flood zones post-deployment while attempting to optimize for low traffic.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI systems are used in critical infrastructure, specifically traffic management and road safety, where failures can lead to life-threatening situations.

AI system and alleged parties

• AI system: Google Maps, Waze (Google)
• AI purpose: Navigation Assistant; Content Recommendation
• Behaviour type: Assistant
• Alleged developer: Google
• Alleged deployer: Google
• Alleged harmed parties: Motorists

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 Negligible, indirect Negligible
• Property: direct Minor, indirect Negligible
• Financial: direct Minor, 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 Minor, indirect Negligible
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Property

Reported: Yes, the reports describe property damage, specifically a vehicle stalling in floodwaters.

Directly caused: A 22-year-old driver's SUV stalled and was potentially damaged in high water in Lumberton, North Carolina, after following GPS directions.

Inferred additional harm: Other vehicles may have suffered minor heat, smoke, or water damage from being routed too close to active fires and floods.

Financial

Reported: Yes, the reports imply financial losses from vehicle damage, though no specific monetary figures are detailed.

Directly caused: Financial loss associated with the repair or replacement of the stalled SUV in North Carolina, though the exact value is not specified.

Inferred additional harm: Inferred financial losses from vehicle repairs for drivers routed into hazardous areas, as well as potential economic losses from delayed emergency responses.

Psychological

Reported: Yes, the reports describe drivers experiencing stress and fear when directed into active fire zones.

Directly caused: Drivers experienced acute stress, anxiety, and fear when their navigation apps directed them into neighborhoods actively on fire or flooded.

Inferred additional harm: Widespread temporary psychological distress and anxiety likely affected dozens of drivers who found themselves unexpectedly facing flames or rising waters due to GPS directions.

Epistemic

Reported: Yes, the apps provided false/misleading routing information by showing fire-threatened and closed roads as clear and safe.

Directly caused: The apps generated misleading navigation routes that failed to reflect active road closures and extreme hazards, leading users to believe the roads were safe.

Inferred additional harm: Thousands of users in the affected areas received inaccurate representations of road safety, temporarily eroding trust in real-time mapping data during emergencies.

People affected

• Occurrences reported: 2
• People reportedly harmed: 1
• People reportedly exposed: 200000

Potential causes

Management

• Prioritization of Speed: Management prioritized travel time savings over safety in emergencies.
• Inadequate Disaster Planning: App companies did not prepare for rapid, unpredictable crises.

Technology

• Traffic-Optimization Bias: Algorithms routed drivers to empty roads which were actually on fire.
• Lack of Hazard Awareness: Routing algorithms had no native capability to detect wildfire zones.
• Real-Time Map Update Lag: Apps failed to instantly reflect fast-moving environmental hazards.

Data Inputs

• Missing Evacuation Data: Algorithms lacked real-time boundaries of active evacuation zones.
• Delayed Road Closure Reports: Official closures took hours to be transmitted and updated on maps.
• Reliance on Crowdsourcing: Crowdsourced reports were too slow during rapidly evolving crises.

Human Factors

• Driver Overreliance on GPS: Users blindly followed app directions despite visible smoke and fire.
• Distracted Driving in Crises: Stressed drivers relied on apps instead of local emergency signs.

Process and Methods

• Slow Verification Checks: Official closure data required multi-step checks before app update.
• Inadequate City Coordination: No automated system existed to share real-time city data with apps.

Regulatory Environment

• No Mapping Liability Standards: Lack of legal framework holding app makers liable for unsafe routing.
• Lack of Emergency Routing Rules: No regulatory mandates forcing apps to integrate disaster hazards.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent, multi-source accounts of navigation apps directing drivers into fire and flood zones due to algorithmic traffic optimization. The technical mechanism (empty roads appearing attractive to the algorithm) is clearly explained across several articles, and statements from both local authorities and Google/Waze confirm the events.
• Ambiguities identified: The exact number of drivers who actually entered dangerous areas solely due to the apps is not precisely quantified, though several specific anecdotes and social media reports are provided.
• Alternative interpretations: None. The causal link between the routing algorithms and the unsafe directions is clear and undisputed.

During the 2017 Southern California wildfires, routing algorithms of popular navigation apps directed drivers into active fire zones by misinterpreting evacuated, empty roads as optimal, low-traffic routes. This caused minor localized physical risks and temporarily obstructed emergency response vehicles, highlighting systemic safety limitations of consumer AI systems operating in disaster zones.

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

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding the integration of commercial navigation AI in emergency management rather than an active, immediate crisis.
• Autonomy: Human-supervised. The navigation apps autonomously calculated and recommended routes, but final driving decisions and physical execution remained with human drivers.
• Novelty: Evolved capability. Represents an evolved capability where standard traffic optimization algorithms failed unexpectedly under extreme, non-standard environmental conditions.

Impact by dimension

• Physical security: Minor. Navigation routing failures directed drivers to hazardous areas and temporarily obstructed emergency response vehicles, representing a minor, localized disruption to emergency infrastructure.
• Information security: Negligible. No significant intelligence compromise or information warfare operations indicated in the incident reports.
• Sovereignty: Negligible. The incident did not threaten state authority, territorial control, or core national government operations.
• Economic security: Negligible. Minor localized property damage (one reported stalled SUV) does not constitute a threat to national economic or technological security.
• Societal stability: Negligible. While drivers experienced temporary stress, there was no systematic threat to civil liberties, human rights, or broad societal stability.