AI liability across the real economy: A sector risk report

Where AI-driven decisions become physical, financial and operational risk

AI is moving beyond software and analysis into machines and systems that navigate, trade, inspect, lift, design and make operational decisions.

As AI assumes a more active role in the physical economy, a single incident may engage several layers of responsibility: the machine, the software controlling it, the organisation operating it and the autonomous decision made at the critical moment.

This sector risk report examines how AI liability is emerging across commodities, financial services, energy, aerospace, construction, shipping and insurance. It is designed for boards, general counsel, risk teams, insurers and operational leaders assessing where AI now sits on their risk register.

When AI enters the physical world

A robot does not need to resemble a person. It may be an automated vehicle moving cargo through a terminal, a subsea machine inspecting an offshore asset, a robotic arm operating on an assembly line or an AI-controlled crane.

When these systems cause injury, property damage or commercial loss, the liability analysis can involve four connected layers:

• The product: the physical machine
• The system: the AI software controlling it
• The operator: the organisation deploying and supervising it
• The decision: the real-time action taken by the AI at the critical moment

The report explores how those layers interact and why conventional allocations of responsibility may be tested by adaptive or autonomous systems.

Seven sectors, different risks, recurring legal questions

Commodities

AI is influencing trading, cargo classification, documentation and terminal operations. Potential disputes may concern abnormal trading patterns, cargo misclassification, mispricing, warranties, demurrage and physical damage arising from automated terminal systems.

Financial services

AI models are being used in areas including credit assessment, trading, fraud detection, customer onboarding and portfolio management. Exposure can arise even where a model is operating as designed, particularly where its limitations, calibration or likely commercial effects have not been adequately anticipated or communicated.

Energy

AI-supported systems are becoming involved in asset monitoring, industrial controls and complex energy infrastructure. Questions following an incident may include whether a system was defective, whether operators relied too heavily on its output and whether testing should have identified the relevant failure mode.

Aerospace

Automated and AI-assisted systems raise difficult questions about human oversight, operational intervention and the division of responsibility among technology providers, service providers and operators. The report considers the lessons that earlier automated systems offer for more adaptive technology.

Construction

AI-assisted modelling, scheduling and structural design can involve architects, engineers, contractors and technology suppliers. If a design is unsafe or non-compliant, responsibility may turn on where AI contributed to the workflow and who reviewed its output.

Shipping

The shift from tools that inform navigational decisions to systems that actively shape them can alter the subsequent analysis of fault. AI-influenced disputes may involve navigation, collisions, cargo, sanctions exposure and the integrity of voyage and positional data.

Insurance

AI-related losses may not fit neatly within traditional products, professional indemnity, technology or cyber policies. Policy wording, risk classification and disclosure are therefore likely to become important areas of dispute for insurers and policyholders.

What the report examines

• How embodied and agentic AI can create physical as well as digital losses
• Where responsibility may sit among developers, hardware providers, operators and users
• Emerging AI dispute scenarios in seven major industries
• The importance of model, control-system and operational records
• Contractual allocation of risk across complex supply chains
• Insurance classification and disclosure issues
• Why AI disputes frequently involve multiple parties and jurisdictions

The report also provides a practical recommendation for each sector and explains why these disputes often need to be treated as cross-border, multi-party liability matters rather than isolated technology claims.

A recurring issue across every industry

The technology may differ, but the underlying questions frequently repeat:

• What was the AI authorised to do?
• Who was responsible for supervising it?
• Could a human intervene?
• Which party controlled the relevant data and configuration?
• Can the system’s actions be reconstructed?
• Do contracts clearly allocate responsibility?
• Which law, jurisdiction or dispute forum applies?
• Does the insurance programme respond?

The report examines how these questions present differently in trading, transport, infrastructure, design, industrial operations and insurance, without assuming that one legal or operational answer fits every sector.