Insurers Exclude Physical AI as Liability Risk Outpaces Coverage

Insurers are excluding physical AI from standard policies, leaving companies that deploy autonomous robots, drones, and vehicles exposed to uninsured losses. The exclusions shift the liability burden onto operators and hardware makers, with no clear backstop for financial damage from AI failures.

Berkshire Hathaway sought state regulatory approval to exclude AI-related damages from general liability policies. Chubb and Travelers filed similar requests. Regulators approved more than 80% of those filings, according to PYMNTS. Some carriers adopted absolute exclusions across multiple policy lines.

When a robot fails, the liability chain is fractured. A hardware maker built the machine. A separate firm supplied the AI model. The operator set the rules. Contracts allocate those duties on paper. A single incident can trigger disputes over which failure caused the loss.

Existing policies leave gaps. A standard business insurance policy may not cover the loss if it includes an AI exclusion or if the loss does not fit the definition of an accident. If it does respond, it likely covers only physical damage, not the production losses that follow when a robot failure shuts down a line.

Two specialty products have emerged. Axis Insurance built a program for companies that make and deploy autonomous robots. The policy covers bodily injury and property damage from AI navigation failures or cyberattacks that take control of a robot. It covers production losses when a software update or sensor failure takes a robot offline, even without physical damage. Defects in third-party components are covered, with the claim landing on the integrator rather than the part maker.

Relm Insurance launched PONTAAI in January, covering bodily injury and property damage among other liabilities.

Neither product covers the full chain. A single robot incident can trigger claims against the hardware maker and the software developer. The operating company may also face claims.

Pricing commercial risk is harder with physical AI. Insurers rely on past claims and equipment records. Physical AI provides less history and more variables. The same robot carries different risk at two warehouses because floor layouts differ and workers interact differently. Software versions and operating limits change after the policy is written.

Lucy Pilko, CEO Americas at AXA XL, noted the pattern matches the early days of cyber insurance, when adoption outpaced understanding of how risk translated into insured loss.

The accumulation problem is severe. Gallagher Re found that product liability coverage may respond to injury and property damage, only in jurisdictions that treat AI software as a product. Financial losses from the same failure get no coverage. A flaw in one widely used AI model can spread instantly to every business running it, across industries and countries.

Berkshire Hathaway (Alpha Score 50, Mixed on AlphaScala's BRK.B stock page) is among the carriers leading the exclusion wave. Axis and Relm have filled some gaps. The full chain of liability remains uninsured in most jurisdictions.