Insurers start writing policies for AI hallucinations and agent errors
TL;DR: A new class of insurance products is emerging to cover losses from AI mistakes, including hallucinations and autonomous agent errors. Some insurers are writing specific “AI malfunction” policies, while others are adding blanket AI exclusion clauses. Deloitte projects the global AI insurance premium market could reach $4.8 billion by 2032, potentially exceeding the cyber insurance market.
The rise of agentic AI, where software bots handle business tasks without constant human supervision, is forcing the insurance industry to confront a new category of risk. When an AI agent independently makes purchasing decisions, generates client communications, or manages inventory, the question of who pays when it goes wrong is no longer theoretical.
Specialist insurer Armilla tests AI models for vulnerabilities before offering coverage and evaluates whether clients meet international risk management standards. Some risks are off the table entirely: Armilla will not cover anything involving medical diagnostics or mental health applications.
The end of “silent coverage”
Until recently, AI-related losses were handled implicitly under existing insurance policies, a situation researchers at Willis Towers Watson have called “silent coverage.” That mirrors the early days of cybercrime insurance, when policies neither explicitly included nor excluded digital risks.
That era is ending. Brokerage firm Founder Shield now builds “AI malfunction and hallucination” scenarios into professional services policies, covering losses that AI systems cause to clients. For additional premiums, coverage can extend beyond digital systems to real-world consequences, such as an AI agent ordering excess inventory.
At the other end of the spectrum, some standard policies now include absolute AI exclusion clauses, refusing any coverage for AI-related incidents. Businesses deploying AI agents need to check their existing policies carefully. One commercial real estate firm discovered its AI agent could not be covered as a regular employee and had to arrange a specialist policy instead.
A market that could outpace cyber insurance
Munich Re, one of the world’s largest reinsurers, provides coverage for both AI model developers and businesses that deploy the technology. Its head of AI insurance, Michael von Gablenz, estimates the AI risk market could grow larger than cybersecurity insurance. Deloitte’s projection of $4.8 billion in global premiums by 2032 gives a sense of the scale.
Looking forward
For UK businesses adopting AI agents, the insurance question deserves attention before deployment rather than after the first incident. The rapid shift from silent coverage to explicit policies, in both directions, means assumptions about existing cover may not hold. Companies should review their policies now and budget for specialist AI coverage if their use cases carry material risk.