Drawn from published evidence and regulatory guidance specific to energy and utilities. Each is pre-scored on a 5×5 likelihood × impact matrix in the Risk Register tool and referenced in the generated policy.
CriticalLikelihood 3 · Impact 5
AI Grid Management System Failure Causing Widespread Power Outage or Cascading Failure
An AI energy management system, grid balancing algorithm, or AI-controlled network switching system exhibits unexpected behaviour — through model drift under novel grid conditions, adversarial manipulation of sensor inputs, or AI response to a combination of simultaneous faults outside its training distribution — causing incorrect switching decisions, failure to activate balancing reserves, or erroneous protective relay actions that trigger cascading disconnections, resulting in large-scale power outages affecting millions of consumers, hospitals, and critical national infrastructure.
CriticalLikelihood 3 · Impact 5
Cyberattack Exploiting AI Vulnerabilities in Energy OT to Sabotage Critical Infrastructure
State or sophisticated non-state threat actors exploit vulnerabilities specific to AI components in energy operational technology networks — including adversarial manipulation of AI sensor data processing, poisoning of AI predictive models through compromised training pipelines, or exploitation of AI decision APIs — to cause unsafe plant conditions, sabotage grid operations, corrupt energy market systems, or trigger physical damage to generation and transmission assets in attacks that conventional OT security controls were not designed to detect or prevent.
HighLikelihood 3 · Impact 4
AI Energy Market Manipulation and Algorithmic Price Distortion Harming Market Integrity
AI energy trading algorithms — whether deployed intentionally to manipulate or exhibiting emergent market-distorting behaviour not intended by their designers — generate artificial price signals, withhold generation capacity at critical system stress moments, engage in coordinated bidding patterns that exploit AI market intelligence advantages, or create feedback loops with competing AI trading systems that amplify price volatility, harming market integrity, increasing consumer energy costs, and attracting REMIT enforcement and FERC market manipulation investigation.
CriticalLikelihood 4 · Impact 5
AI Demand Forecasting Error Creating Grid Instability or Insufficient Reserve Margins
AI electricity demand forecasting, renewable generation forecasting, and grid balancing systems produce materially inaccurate predictions — through model failure during unprecedented weather events, failure to account for rapid demand-side changes, or systematic bias against low-frequency high-impact scenarios — leading network operators to procure insufficient reserve capacity, fail to activate demand response at critical moments, or make incorrect interconnector scheduling decisions that create supply-demand imbalance threatening grid stability.
HighLikelihood 4 · Impact 4
AI Smart Metering and Dynamic Tariff Systems Causing Consumer Harm and Fuel Poverty
AI-driven dynamic energy tariff optimisation, smart meter-based billing AI, and AI demand response systems that adjust consumer prices and supply in real time produce harmful outcomes for vulnerable consumers — including elderly customers on fixed incomes, households with medical devices requiring continuous power, and low-income households — through incorrect billing, unexpected price spikes communicated without adequate notice, AI-directed disconnection of vulnerable consumers, or demand response signals that deprive vulnerable households of heating or cooling during extreme weather.
CriticalLikelihood 2 · Impact 5
AI in Nuclear and Chemical Facility Operations Creating Safety Case Invalidation
AI systems deployed in nuclear power plant operations, process control of chemical manufacturing facilities, or management of hazardous industrial processes are not adequately reflected in the safety case submitted to the nuclear regulator or chemical safety authority — either because AI was added to existing processes without safety case re-evaluation or because AI adaptive behaviour creates plant states that were not modelled in the safety assessment — invalidating the regulatory basis for continued operation and creating potential for uncontrolled hazardous events.