How insurers are planning coverage for businesses using AI
What's the story
As businesses increasingly rely on artificial intelligence (AI) "agents" to independently drive revenue growth, some insurance companies are stepping in to cover potential mistakes. However, others are still hesitant. Phil Dawson, head of AI policy and partnerships at specialist insurer Armilla, explained that the goal of advanced AI is to significantly reduce human involvement in decision-making processes.
AI evolution
AI's impact on insurance
The emergence of "agentic AI" has seen bots autonomously perform computer tasks, leading to a reduction in human workforce. Dawson noted that this trend poses challenges to the fundamental principles of existing insurance coverage. Despite their best efforts, companies in the AI race have not completely eliminated risks such as "hallucinations," where false information is presented with confidence.
Insurance response
'Silent coverage' for AI risks
AI-related liability risks have been implicitly covered under insurance policies, a practice known as "silent coverage," according to analyst Sonal Madhok and law professor Anat Lior. However, they argue that this is similar to the liability coverage questions seen in the early days of cybercrime. They predict that we will soon see policies explicitly addressing AI, marking an end to the silent coverage era.
Policy changes
Insurers' response to AI-related incidents
Insurers are moving beyond their "wait-and-see approach" toward AI-related incidents, Jonathan Mitchell of brokerage firm Founder Shield said. Some standard insurance policies now have "absolute AI exclusion" clauses that deny coverage for AI-related incidents. Dawson cited an example of a commercial real estate company that tried to cover its AI agent as a regular employee but had to switch to a special policy.
Coverage expansion
Specialized coverage for AI malfunctions
Founder Shield offers specialized coverage for "AI malfunction and hallucination" scenarios under professional services policies. These policies cover losses caused by the technology to clients. For an additional cost, the coverage can be extended beyond computer networks to real-world harm, such as AI mistakenly ordering excessive inventory for a company.
Risk evaluation
Testing and limitations of coverage
Armilla tests AI models for vulnerabilities before offering coverage and checks if the client's risk management framework meets international standards. However, like other insurers, Armilla can refuse certain risks. For instance, it does not offer coverage for anything related to medical diagnostics or mental health-focused applications.
Market growth
Munich Re's perspective on AI insurance market potential
Munich Re, a global leader in insurance and reinsurance, offers coverage for companies designing as well as using AI models. Michael von Gablenz, head of AI insurance at Munich Re, acknowledged the risk of models making errors or hallucinating cannot be completely avoided. He stressed that AI systems are statistical models; and any statistical model has uncertainty in it. Despite this risk, the potential market for AI insurance is huge.