Starship Delivery Robot Ran into Problems Traversing Campus Terrains

A Starship autonomous delivery robot struggled to navigate campus terrains of UCLA, reportedly getting stuck into a planter and falling off the stairs.

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 normal sidewalk conditions, slipping into a planter and later hitting a scooter, demonstrating a lack of capability and robustness in navigating real-world obstacles.

Causal factors

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

The incident was initiated by the autonomous navigation system of the Starship robot failing to maintain its path, leading to an unexpected getting-stuck event during real-world deployment.

EU AI Act risk tier

• Risk tier: 4 Minimal or No Risk

Minimal or No Risk: The delivery robot is a simple application used for food delivery, posing low or negligible risks to users and society under the EU AI Act.

AI system and alleged parties

• AI system: Starship delivery robot (Starship Technologies)
• AI purpose: Autonomous Driving; Navigation Assistant
• Behaviour type: Hybrid
• Alleged developer: Starship Technologies
• Alleged deployer: Starship Technologies
• Alleged harmed parties: Starship Technologies

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

People affected

• Occurrences reported: 1
• People reportedly exposed: 2

Potential causes

Management

• Premature Public Deployment: Deploying in complex campus environments before solving edge cases.

Technology

• Blind Spots in Camera Angle: Low camera angle prevented remote operator from seeing the stairs.
• Inadequate Off-Road Recovery: Robot could not autonomously clear the path lip after slipping.
• Turn Signal Deactivation: Turn signals disabled during remote control, reducing predictability.

Data Inputs

• Misinterpreted Physical Input: Skaters lifting the robot triggered a false vandalism alert.
• Incomplete Obstacle Detection: Sensors failed to detect the drop-off at the edge of the path.

Human Factors

• Operator Situational Awareness: Remote operator drove robot down stairs due to limited visibility.
• Well-Intentioned Interference: Pedestrians physically moved the robot, triggering remote takeover.

Process and Methods

• Path-Following Policy: Routing algorithm forced the robot to the extreme right edge.
• Remote Handover Protocol: Vandalism alert immediately handed control to operator without context.

Regulatory Environment

• Lack of Public Pathway Rules: No standard rules for autonomous robots sharing pedestrian paths.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides a highly detailed, first-hand chronological log of the event, supported by video observations and context from the developer's public statements. The sequence of events is clear and unambiguous.
• Ambiguities identified: The exact extent of damage to the scooter and the robot is not specified, nor is the identity of the scooter's owner.
• Alternative interpretations: The incident could be interpreted primarily as a human remote-operator error rather than an AI failure, since the most dramatic event (jumping the stairs) occurred under manual control.

A commercial food delivery robot experienced localized navigation and remote-control failures at UCLA. The incident resulted in minor property damage with zero national security implications.

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

Threat characteristics

• Imminence: Long-term. Represents an isolated, localized navigation failure with no active national security crisis or developing threat.
• Autonomy: Human-supervised. The robot operates autonomously but is supervised and can be overridden by remote human operators, as occurred during the incident.
• Novelty: Established threat. Minor navigation failures and remote operator control issues in commercial delivery robots are established technical limitations.

Impact by dimension

• Physical security: Negligible. Minimal physical impact restricted to minor property damage of a scooter and the robot itself on a university campus. No threat to critical infrastructure or human safety.
• Information security: Negligible. No involvement of information warfare, intelligence compromise, or systemic disinformation campaigns.
• Sovereignty: Negligible. No impact on state authority, government decision-making, or sovereign functions.
• Economic security: Negligible. Minor localized economic impact including potential repair costs, but no threat to strategic industries or national economic security.
• Societal stability: Negligible. Localized incident with no threat to societal stability, civil liberties, or population-scale safety.