FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Insurance wiki
Agentic AI liability is the legal responsibility for loss caused when an AI system pursues a goal and takes actions with limited or no human input. It matters because the AI cannot be sued, so loss falls on those who build, supply, deploy or use it, and autonomy makes cause harder to trace.
Part of: AI insurance UK
In short
Agentic AI acts as well as answers: it can send messages, make payments or change code. In English law the AI has no legal personality, so liability for its actions falls on developers, deployers and users. The Law Commission's July 2025 discussion paper flags liability gaps, causation and opacity as key issues, without proposing reform. The UKJT's July 2026 statement says existing law copes and that deployers are likely to be liable for autonomous AI harm unless it was unforeseeable. Insurers place agent risk into PI, cyber, liability and IP cover, subject to the policy terms.
Agentic AI is AI given an objective plus the tools and permissions to pursue it: it can send messages, update records, make payments or change code. Agentic AI liability is who answers in law when those actions cause loss.
The Law Commission's discussion paper Artificial Intelligence and the Law (31 July 2025) defines autonomy as "the ability of an AI system to complete objectives with limited or no human input, control or oversight" (p. 8). Agentic systems sit at the autonomous end. A chatbot that just talks raises statement risk (see AI chatbot insurance); an agent that acts raises action risk (see AI agents insurance).
The Law Commission paper raises these issues but, in its words, "does not contain within it proposals for reform" (Law Commission paper, p. 3).
| Issue | What the Law Commission paper says | Practical effect |
|---|---|---|
| Legal personality | AI systems "do not currently have separate legal personality and therefore can neither be sued or prosecuted" (p. 9) | Loss must be pinned on a person or company |
| Liability gaps | "Liability gaps", where no person is liable, are "more likely to occur in connection with highly autonomous and adaptive systems" (p. 9) | Claimants look to deployers and contracts |
| Causation | A defendant might argue an unpredicted output was an intervening cause (p. 10), yet "with AI, unexpected outcomes are expected" (p. 11) | "The agent did something unexpected" may not break the chain of causation |
| Supply chain | AI supply chains "can therefore be intricate and complicated" (p. 13) | Several defendants. See apportionment between defendants |
| Opacity | It can be "exceedingly difficult" to establish why a system produced an output (pp. 14 to 15) | Logs and testing records become key evidence |
"Agent" is a product label, not a legal status. The UK Jurisdiction Taskforce's Legal Statement on Liability for AI Harms (July 2026), as summarised by practitioners, concludes that:
The statement is not legally binding, and contract formation by AI is outside its scope. For cases, see AI agent insurance.
One autonomy-specific liability rule already exists. Under section 2 of the Automated and Electric Vehicles Act 2018, in force since 21 April 2021, where an accident is caused by an automated vehicle "when driving itself" on a road or other public place in Great Britain and the vehicle is insured, the insurer is liable for the resulting damage, and section 5 lets it recover from anyone else liable. No equivalent general rule exists for AI agents. See the Automated and Electric Vehicles Act 2018 and robotics and autonomous systems insurance.
Agent risk usually sits in existing covers. Examples are illustrative; cover is subject to the policy terms.
| Agent loss | Cover most likely to respond |
|---|---|
| A wrong action causes a client financial loss | Professional indemnity |
| The agent leaks data or is hijacked through prompt injection | Cyber. See cyber and AI risks |
| The agent controls equipment that injures someone or damages property | Product or public liability. See AI product and public liability |
| The agent publishes content that infringes someone's rights | IP infringement cover |
Underwriters usually ask what each agent may do without human approval, how it is tested, monitored, logged and stopped, and what its contracts say. A Lloyd's Market Association wordings forum presentation in April 2026 suggested proposal questions such as who checks AI output. Because some wordings now carry AI exclusions, agent work should be named in the policy as insured activity, not left silent. See AI insurance UK.
It is legal responsibility for loss caused when an AI system takes actions, such as sending messages, making payments or changing code, with limited human input. The AI has no legal personality, so liability falls on developers, deployers, users and others in the chain.
Usually the business that deployed it, towards its customers and clients, with any recovery from the vendor depending on contract. The July 2026 UKJT statement considers those who deploy autonomous AI likely to be liable for the harm it causes unless that harm was unforeseeable.
No. In English law AI systems have no separate legal personality, so they cannot be sued or prosecuted. The Law Commission's July 2025 discussion paper said it is not clear that any AI system is advanced enough to warrant legal personality.
It can, through existing covers: PI for clients' financial loss, cyber for data and security events, product or public liability for injury or damage, and IP infringement cover, subject to the policy terms. Check for AI exclusions and name the agent work as insured activity.
Tell us what your AI work is, where you are based, your turnover and the limit you need. It takes a few minutes online, and a broker will come back to you with options. Or call 0117 325 0027.
Start an AI insurance quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information, not advice on your individual circumstances, and it does not guarantee that cover will be available or on what terms.
