Protecting against malfunction
As robots begin to move autonomously in the world and interact with humans in free space at scale, more accidents will happen
We’re moving beyond industrial robotics that are generally stationary, confined arms to robots that are autonomously moving around the world in free space. These range from autonomous vehicles to humanoids to delivery robots to warehouse & logistics to retail to healthcare assistants to surgeons. Eg. Intuitive Surgical set aside $67M to settle 3,000 da Vinci-robot injury claims.
This will cause accidents as proportional to robots deployed which should scale exponentially. In a world where robotics becomes a several trillion-dollar industry, chasing a slice of that would make for a very valuable ground for a startup.
This paradigm will also change the nature of what insurance companies are diligencing or estimating annuities against. It’s effectively a form of product liability insurance that provides liability protection against manufacturing errors, AI model mistakes, software glitches, unplanned downtime, hacks, etc. Geico (and now Berkshire) CEO even recently spoke about how AVs will force them to shift their auto insurance practice from underwriting human error to product malfunction.
Maybe the issue of systemic loss is resolved by reinsurers imposing a standard cap of losses as condition for coverage of the issuing policymaker.
Incumbents have no entrenched advantage in underwriting these questions because these systems are just starting to be deployed so there’s no existing datasets. Startups may actually be far better at underwriting because diagnosing and predicting a robot’s likelihood of malfunction requires deep technical knowledge of how both their software and hardware works. Maybe an insurance startup focused on working with robotics startups early in deployment could get sub-second telemetry, updates OTA firmware, fleet-wide anomaly detection.
In fact, incumbents are retreating from offering coverage of AI entirely.
A startup could even use computable contracts to fully automate their contract execution and claims processing.
You could train a system to:
The absolute deterministic execution of both computable contracts and formal verification may be the only way to fully automate and deploy complex agentic systems at scale in such highly regulated, high downside industries. Otherwise each autonomous transaction may require manual checking. A startup may even have a shot at fundamentally disrupting their business as opposed to remaining an innovation supplier with these tools and by going after completely new, greenfield markets like robot insurance.
That’s how to build the $100B insurance company of the 21st Century.
https://www.ft.com/content/abfe9741-f438-4ed6-a673-075ec177dc62
https://eprints.whiterose.ac.uk/id/eprint/217393/
https://www.drugwatch.com/davinci-surgery/lawsuits/?utm_source=chatgpt.com
