Thinking on frontier risk
Welcome to our journal. We share case studies, research findings, and our perspective with fellow insurance professionals, and we're always glad to collaborate on future publications. For inquiries, contact us.
How AI liability is underwritten
AI liability is priced from the deployer's duty, not the vendor's product. Authority, containment, and compliance, read from an application and a public record. No audit, and nothing installed to price the risk.
Why no policy you own responds when your AI does the wrong thing
Cyber pays when the system stops working. This pays when the system works perfectly and does the wrong thing. A walk through the tower a buyer already holds, and where each line stops.
You can't insure what you can't describe
Deployed AI agents take real actions with real consequences. Before that risk can be underwritten, it has to be made legible, which is what certification is for.
The gap isn't in the AI. It's in the oversight.
Standard professional liability policies are quietly adding AI exclusions. The exposure they leave behind isn't a machine failure, it's a human one.