Before the first claim: A General Counsel’s guide to AI liability risk

Responsibility will depend on the facts, contracts and applicable law. The deploying entity may face the most immediate exposure because it placed the system into operation and relied on its outputs. However, responsibility may also extend across developers, data providers, integrators, testers, maintenance providers and other participants. The guide explains how to map that chain before a claim arises.

Yes. A dispute may concern a system that performed its intended function but produced harmful results in circumstances its designers or deployers did not adequately anticipate. The relevant questions may therefore concern system design, operational limits, oversight and safeguards rather than a conventional coding error.

Relevant material may include system logs, model-version records, configuration data, design documentation and records concerning human overrides or emergency intervention. The exact preservation exercise should reflect the system, its use and the possible claims. The full guide provides the practical framework.

AI systems may depend on multiple contracts covering data, software, APIs, integration, licensing, maintenance and operational deployment. Responsibility and dispute-resolution provisions may differ across those agreements, making it difficult to pursue all potentially responsible parties in one forum.

The answer may not be straightforward. Depending on the circumstances, arguments may arise over products liability, professional indemnity, technology errors and omissions, cyber or other coverage. Disclosure of material AI deployments to insurers may also become important. The guide examines the questions organisations should raise with their brokers and insurance teams.