Uber AV Killed Pedestrian in Arizona

An Uber autonomous vehicle (AV) in autonomous mode struck and killed a pedestrian in Tempe, Arizona.

Risk classification

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

The primary failure was the autonomous vehicle's perception system failing to correctly classify a pedestrian and predict her path, representing a critical lack of capability and robustness in real-world conditions.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The safety driver exhibited automation complacency, a known effect of overreliance, by streaming a video instead of keeping eyes on the road.
• 6.5 Governance failure: Arizona officials established a low-regulation environment to attract self-driving companies, failing to provide adequate safety oversight.

Causal factors

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

The incident was caused by the Uber autonomous vehicle's perception system failing to correctly classify a pedestrian and take evasive action during public road testing. This unexpected failure occurred post-deployment during real-world operations.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is an autonomous vehicle operating in public transport infrastructure. The report describes a 'semi-autonomous Uber vehicle' and 'fully autonomous vehicle technology' which falls under AI in critical infrastructure and safety-critical transport applications, posing significant implications for safety and public interest.

AI system and alleged parties

• AI system: Uber autonomous vehicle
• AI purpose: Autonomous Driving; Identification
• Behaviour type: Autonomous
• Alleged developer: Uber
• Alleged deployer: Uber
• Alleged harmed parties: pedestrians, Elaine Herzberg

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. 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 Substantial, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Minor, indirect Negligible
• Financial: direct Substantial, indirect Substantial
• 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 Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly describes a fatality resulting from the collision.

Directly caused: Elaine Herzberg, a 49-year-old pedestrian, was struck by the vehicle at 39 mph and died from multiple blunt-force injuries.

Property

Reported: The report explicitly describes damage to the vehicle and the victim's bicycle.

Directly caused: The victim's bicycle was smashed and scraped along the pavement, and the Uber Volvo SUV sustained front-end damage.

Financial

Reported: The report explicitly describes financial settlements paid by Uber to the victim's family.

Directly caused: Uber paid settlements in the low millions of dollars to Elaine Herzberg's husband, daughter, parents, and son.

Indirectly caused: Uber benched and eventually laid off nearly 300 employees in Tempe, resulting in job losses and severance costs.

Inferred additional harm: Uber likely incurred significant financial losses from suspending its autonomous testing program, legal defense fees, and the eventual sale/restructuring of its Advanced Technologies Group.

Psychological

Reported: The report explicitly describes severe psychological distress and trauma experienced by the safety driver and the victim's family.

Directly caused: The safety driver, Rafaela Vasquez, experienced severe trauma, anxiety, and spiraled into agoraphobia and isolation following the crash and subsequent public backlash.

Indirectly caused: Elaine Herzberg's daughter and parents experienced severe grief and distress over her death.

People affected

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

Potential causes

Management

• Prioritizing Miles Over Safety: Uber pressured operators to 'crush miles' to show rapid progress.
• Inadequate Safety Culture: NTSB found Uber did not make safety a top priority in its operations.
• Ignoring Internal Warnings: Executives ignored manager warnings about poor training and crash risks.

Technology

• Unstable Object Classification: System constantly reclassified the pedestrian, resetting path prediction.
• Disabled Volvo Emergency Braking: Uber disabled Volvo's factory AEB to prevent system conflicts.
• Braking Suppression Logic: System delayed hard braking to prevent erratic, false-alarm stops.

Data Inputs

• Lack of Jaywalker Labeling: System was not trained to recognize pedestrians outside of crosswalks.
• Confusing Sensor Data: Victim carried a bike with bags, which confused classification models.
• Poor Night Lighting: The crash occurred at night on a poorly illuminated four-lane road.

Human Factors

• Automation Complacency: Safety driver trusted the automation and stopped paying active attention.
• Driver Distraction: Safety driver was streaming a TV show on her personal phone while driving.
• Delayed Human Intervention: Driver took control only two-hundredths of a second before the impact.

Process and Methods

• Transition to Solo Operators: Uber cut the second safety driver, increasing the chance of distraction.
• Inadequate Driver Monitoring: Management did not perform real-time monitoring or frequent video checks.
• Lack of System Transparency: Proprietary software made it difficult to audit and verify safety limits.

Regulatory Environment

• Lax State Oversight: Arizona established a regulation-free zone to attract tech companies.
• Permissive Testing Approvals: State allowed testing on public roads without strict safety verification.
• Weak Federal Standards: Federal regulators relied on voluntary safety reporting from companies.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide exceptionally detailed, multi-perspective accounts of the incident, including technical details from the NTSB investigation, personal backgrounds of the operator and victim, and the regulatory context. The facts of the collision, the system's classification failures, and the operator's distraction are thoroughly documented and consistent across sources.
• Ambiguities identified: None that significantly affect the classification of the primary safety failure or the resulting harms.
• Alternative interpretations: One could argue the primary failure was human negligence (the distracted safety driver), but the NTSB and technical reports make it clear that the AI system's perception design flaws and the disabling of the emergency braking system were critical causal factors.

This incident involved a fatal collision where an Uber autonomous vehicle struck and killed a pedestrian due to perception system failures and safety driver distraction. While a significant milestone in AI safety and regulatory history, it remains a localized traffic accident with negligible national security implications.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: No clear primary
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The incident represents an isolated, post-deployment safety failure rather than an active national security crisis.
• Autonomy: Human-supervised. The vehicle was operating in autonomous mode but was under the supervision of a backup safety driver who was expected to intervene.
• Novelty: First-of-its-kind. This incident marked the first recorded pedestrian death caused by an autonomous vehicle, establishing a significant safety precedent.

Impact by dimension

• Physical security: Minor. A single civilian fatality occurred due to an AI perception failure, representing a physical safety issue but carrying negligible direct national security impact. It was managed through standard local law enforcement and federal safety investigation procedures.
• Information security: Negligible. No information warfare, intelligence compromise, or systematic information manipulation occurred during this incident.
• Sovereignty: Negligible. No threats to state authority, territorial control, or core government operations were present in this incident.
• Economic security: Negligible. The economic impact was localized to Uber's corporate operations, settlements, and testing program suspension, with no threat to national economic stability or strategic technological security.
• Societal stability: Negligible. No systematic threats to civil liberties, mass surveillance, or population-scale social manipulation occurred.