MediumEnvironmental Impact●Realized
Compute-Driven Carbon Footprint
Societal & Economic RiskDescription
Large-scale agent operations consume compute and energy at rates exceeding human-operated systems. AI data-centre energy consumption creates significant environmental impact; water consumption for cooling at scale.
Example scenario
Enterprise deploying 10,000 AI agents to replace email processing increases data-centre energy consumption by 40%.
Real-world evidence●Realized
Air Canada's chatbot provided a passenger with incorrect information about bereavement fare policies, and a Canadian tribunal held the airline legally liable for the chatbot's misleading output in 2024, establishing a confirmed production case of an AI system operating outside its permitted advisory perimeter with regulatory and legal consequences.
Primary mitigations
- Compute efficiency monitoring
- energy consumption targets
- carbon-aware compute scheduling
- hardware efficiency standards.
Detection signals
Energy consumption per agent task; carbon intensity; PUE metrics.
Mitigating controls
4 Dual coverage