Bus Passengers Reportedly Injured After Starship Delivery Robot Entered a Crossing Against a Red Light in Tallinn, Estonia

A Starship Technologies delivery robot misread a traffic light and entered a pedestrian crossing on a red light in Tallinn, causing a bus to brake sharply and injure up to nine passengers.

A Starship delivery robot reportedly entered a crossing against a red light in Tallinn, Estonia. A city bus braked sharply and collided with it, causing passengers to fall; the transit operator reported up to nine injuries. Starship said lighting conditions had affected the robot's traffic-light recognition and announced software changes.

Source: AI Incident Database

Risk classification

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

The robot failed to perform reliably under challenging lighting conditions, misinterpreting a traffic light due to an illuminated advertisement and incorrect light sensitivity settings.

Causal factors

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

The incident was caused by the autonomous decision of the delivery robot's AI system to cross the street on a red light, which was an unexpected and unintended failure of its perception system.

EU AI Act risk tier

  • Risk tier: 2 High Risk

High Risk: The AI system is deployed in physical public infrastructure and transport environments, where failures directly impact physical safety and public interest.

AI system and alleged parties

  • AI system: Starship delivery robot (Starship Technologies)
  • AI purpose: Autonomous Driving; Navigation Assistant
  • Behaviour type: Autonomous
  • Alleged developer: Starship Technologies, Delivery robot developers, Autonomous mobile robot developers
  • Alleged deployer: Starship Technologies, Delivery robot operators, Autonomous mobile robot operators
  • Alleged harmed parties: Transit systems, Tallinna Linnatransport, Road users, Passengers, Bus passengers

Harm severity

Severity ratings are based on this scale.

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

Psychological

Reported: The report explicitly describes psychological distress experienced by the bus driver.

Indirectly caused: The bus driver experienced significant emotional distress due to the near-miss and passenger injuries, preventing him from completing his working day.

Inferred additional harm: Some of the injured or falling passengers likely experienced temporary shock or anxiety from the sudden braking and collision.

People affected

  • Occurrences reported: 1
  • People reportedly harmed: 10
  • People reportedly exposed: 15

Potential causes

Management

  • Acceptance of Residual Traffic Risk: Management accepted that not all traffic risks can be ruled out.
  • Prioritizing Economic Scale: Constant monitoring was omitted to keep the business model sustainable.

Technology

  • Flawed Light Sensitivity: The robot's light sensitivity was not working correctly and had changed.
  • Traffic Light Misinterpretation: Robot misread the traffic light due to line of sight alignment.
  • Autonomous Decision Error: Robot decided to cross because no cars were moving towards it.

Data Inputs

  • Visual Noise from Advertisement: An illuminated advertisement behind the traffic light caused visual confusion.
  • Poor Lighting and Angle: Darkness and the robot's positioning angle affected sensor inputs.

Human Factors

  • Lack of Constant Monitoring: Teleoperators do not constantly monitor robots due to economic constraints.

Process and Methods

  • Post-Accident Protocol Failure: Back-office moved robot to curb instead of waiting for police.
  • Inadequate Real-Time Intervention: Remote operators did not intervene before the robot entered the crossing.

Regulatory Environment

  • Unclear Emergency Reporting: Ambiguity on when remote operators must notify emergency services.

Information quality

  • Classification confidence: High
  • Reason for confidence: The reports provide consistent, detailed accounts of the collision, the technical reasons for the robot's failure from the manufacturer's head of operations, and the resulting passenger injuries.
  • Ambiguities identified: There is a minor dispute regarding whether the robot 'fled' the scene or moved to a safe location to wait by the curb according to safety protocols.
  • Alternative interpretations: None, the facts of the traffic accident and the AI's perception failure are clear.

A Starship Technologies delivery robot's perception failure led to a traffic incident in Estonia, causing minor injuries to nine bus passengers. The event represents a localized public safety issue 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 is a resolved traffic accident, representing a long-term safety and regulatory concern rather than an active national security crisis.
  • Autonomy: Full autonomy. The delivery robot was operating autonomously to navigate streets and make crossing decisions without direct human intervention at the time of the incident.
  • Novelty: Established threat. Sensor perception failures and accidents involving autonomous ground vehicles and delivery robots are established risks with existing precedents.

Impact by dimension

  • Physical security: Minor. A delivery robot's navigation failure caused a minor traffic accident, injuring up to nine bus passengers. This represents a localized safety incident rather than a systemic threat to critical infrastructure.
  • Information security: Negligible. No elements of information warfare, intelligence compromise, or systematic disinformation were present in this incident.
  • Sovereignty: Negligible. This was a localized commercial delivery robot accident with no impact on state authority, government operations, or constitutional processes.
  • Economic security: Negligible. The incident involved minor property damage and localized insurance claims, posing no threat to strategic national economic or technological security.
  • Societal stability: Negligible. The accident did not involve mass surveillance, algorithmic oppression, or threats to civil liberties and societal stability.
Explore in the interactive Incident Tracker