Warehouse robot ruptures can of bear spray and injures workers

An automated warehouse robot at an Amazon facility in Robbinsville, New Jersey, punctured a can of bear repellent, releasing concentrated capsaicin fumes. The incident resulted in 54 workers suffering from respiratory distress and eye/throat irritation, with 24 individuals requiring hospitalization, including one in critical condition. The event has prompted investigations by OSHA and Amazon, and has drawn criticism regarding warehouse safety standards and the risks associated with automated systems.

Twenty-four Amazon workers in New Jersey were hospitalized after a robot punctured a can of bear repellent spray in a warehouse.

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 incident was caused by a physical failure of the warehouse robot, which lacked the capability or robustness to safely navigate around or handle a fallen aerosol can, resulting in a puncture.

Additional risk subdomains

  • 6.2 Increased inequality and decline in employment quality: The incident highlighted union concerns that aggressive automation and productivity demands at Amazon warehouses compromise worker safety and degrade working conditions.

Causal factors

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

The physical harm was caused by an autonomous warehouse robot accidentally puncturing a can of bear repellent during its normal post-deployment operations.

EU AI Act risk tier

  • Risk tier: 2 High Risk

High Risk: The system is an AI-driven worker management and industrial robotics system operating in a workplace environment, which directly impacts the physical safety and health of workers.

AI system and alleged parties

  • AI system: None named
  • AI purpose: Robotic Manipulation; Workforce Administration
  • Behaviour type: Autonomous
  • Alleged developer: Amazon
  • Alleged deployer: Amazon
  • Alleged harmed parties: Warehouse Workers

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 Minor, 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

Physical

Reported: The report explicitly describes physical injuries to dozens of workers due to capsaicin exposure.

Directly caused: Fifty-four workers suffered respiratory distress, burning eyes, and burning throats. Twenty-four workers were hospitalized, including one who was placed in the intensive care unit in critical condition.

People affected

  • Occurrences reported: 1
  • People reportedly harmed: 54
  • People reportedly exposed: 80

Potential causes

Management

  • Productivity prioritized over safety: Pressure to move products quickly during holiday rush compromised safety.
  • Inadequate hazard risk planning: Management failed to address risks of robots handling aerosol repellents.

Technology

  • Robot collision with fallen item: Automated machine failed to detect and avoid bear spray on the floor.
  • Lack of robotic manipulation hands: Robots cannot handle delicate or hazardous items safely.

Data Inputs

  • Inadequate object recognition: Robot sensors failed to identify the fallen aerosol can as a hazard.
  • Lack of real-time telemetry: Automated systems did not alert when the can fell from the shelf.

Human Factors

  • Co-working in automated zones: Humans working in close proximity to automated machines without safety barriers.

Process and Methods

  • Unsafe storage of hazardous items: Aerosol cans stored on open shelves where they could fall into robot paths.
  • Flawed packaging design: Bear spray packaging used flimsy clamshells prone to opening when dropped.
  • HVAC dispersion of fumes: Warehouse ventilation system spread the toxic fumes across multiple floors.

Regulatory Environment

  • Inadequate automation rules: Lack of specific regulatory standards for robot-human co-working safety.
  • OSHA reporting failures: History of failing to report work-related illnesses and injuries.

Information quality

  • Classification confidence: High
  • Reason for confidence: Multiple reputable news sources provide consistent, detailed accounts of the incident, including the specific number of injuries, hospitalizations, the cause of the puncture, and the chemical involved.
  • Ambiguities identified: The exact mechanical action of the robot (whether it ran over the can or dropped it) is slightly inconsistent across reports, though most agree it punctured a fallen can.
  • Alternative interpretations: The incident could be viewed purely as an industrial workplace accident rather than an AI safety failure, but the central role of the autonomous robot makes the AI safety classification highly appropriate.

An autonomous warehouse robot at an Amazon facility accidentally punctured a can of bear repellent, injuring several workers. The incident represents a localized workplace safety issue rather than a national security threat, with negligible impact on critical infrastructure, sovereignty, or national economic security.

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

Threat characteristics

  • Imminence: Long-term. The event was a past, localized workplace accident with no ongoing or imminent threat to national security.
  • Autonomy: Full autonomy. The warehouse robot executed physical tasks autonomously based on environmental inputs without direct human oversight.
  • Novelty: Established threat. Workplace accidents involving industrial automation and physical failures are established risks with existing regulatory frameworks.

Impact by dimension

  • Physical security: Negligible. Localized industrial accident confined to a single commercial facility. It did not affect critical national infrastructure, military systems, or public physical security.
  • Information security: Negligible. No espionage, intelligence compromise, or information operations were associated with this localized physical accident.
  • Sovereignty: Negligible. The incident did not involve or impact state authority, electoral processes, or government operations.
  • Economic security: Negligible. While causing temporary localized operational disruption and minor financial impact to a private company, the incident posed no systemic threat to national economic or technological security.
  • Societal stability: Negligible. The incident was a workplace safety failure impacting a localized group of employees, with no systemic implications for civil liberties or societal stability.
Explore in the interactive Incident Tracker