Liability for autonomous AI harms is unresolved and no reform vehicle exists

The Law Commission's discussion paper (31 July 2025) identifies genuine liability gaps (scenarios where 'no natural or legal person is liable' for harm caused by adaptive autonomous systems), plus causation and opacity problems, but it is a scoping paper with no commissioned reform project behind it. The UK Jurisdiction Taskforce consulted on a draft legal statement on AI liability under English private law (Jan 2026), which can clarify but not change law. Meanwhile AISI finds agent task-complexity doubling roughly every seven-to-eight months: agentic deployment is accelerating ahead of any framework, and insurers cannot price the risk.

Unclear liability chills legitimate deployment (buyers cannot allocate risk) while under-protecting victims (harms with no defendant). As agents transact, hire and advise autonomously, the first major uncompensated AI harm will force rushed, badly-designed legislation: the worst way to make private law.

Commission a full Law Commission reform project with a reporting deadline, leading to an AI Liability Act: attribution rules for agentic systems, strict-liability channels for defined high-risk deployments, mandatory insurance for operators; interim statutory guidance for courts building on the UKJT statement.

// State-led: Instrument: commissioned Law Commission project leading to an AI Liability Act.

Autonomous systems can already cause harm no one is liable for, but the fix needs a multi-year Law Commission project not yet commissioned, and no date forces it.