The companies building AI faster than anyone else just lost their safety net.
The Summary
- AI-related lawsuits in the US jumped 978% from 2021 to 2025, and major insurers (Berkshire Hathaway, Chubb, Travelers) are now adding exclusion clauses that drop AI liability coverage from standard commercial policies.
- The exclusions cover AI-driven discrimination claims, IP violations from training data, and physical damage from autonomous systems.
- Small specialty insurers are stepping in with AI-specific policies, but the retreat of major carriers signals a market repricing of AI risk that most companies haven't budgeted for.
The Signal
The insurance industry just told us something the tech press won't: AI deployment has outpaced our ability to price its actual risk. When Berkshire Hathaway walks away from a line of business, it's not because they're scared of innovation. It's because the math stopped working.
A 978% increase in AI-related litigation over four years isn't noise. It's a pattern. And major commercial insurers have seen enough of the pattern to know they can't underwrite it at current premium rates. The exclusion clauses they're adding cover three buckets: employment discrimination (your AI recruiting tool screens out protected classes), intellectual property theft (your model trained on copyrighted data you didn't license), and physical damage (your autonomous forklift runs into something expensive).
"This is highlighting a crucial blind spot for businesses. They are clamoring to join the AI bandwagon, but they have to pause and ask if they're fully protected."
Here's what matters: these aren't hypothetical risks. Every major tech company has faced copyright claims over training data. Multiple companies have pulled AI recruiting tools after discrimination allegations. Autonomous vehicle incidents are documented daily. The difference is that until now, companies assumed their standard commercial general liability policies would cover the fallout. They won't.
The repricing happens in two ways:
- Existing policies get amended with AI exclusions at renewal
- New AI-specific policies cost significantly more and cover significantly less
- Companies that deployed AI thinking it was "just software" now face uninsured liability
The market response is predictable but revealing. Specialty insurers like HSB are offering AI liability products for small businesses. Smaller, newer carriers are building AI-focused books of business. But this is the insurance equivalent of subprime lending. The majors won't touch it because they've run the actuarial tables. The specialists will, but at prices that reflect true risk, not the subsidized rates companies got when AI liability was bundled into standard coverage nobody was paying attention to.
For companies already deep into AI deployment, this creates a sudden accounting problem. That generative AI customer service bot you launched? The autonomous inventory system in your warehouse? The AI-powered hiring platform? You just went from insured to uninsured, or from affordably insured to expensively underinsured. CFOs are about to have uncomfortable conversations with CIOs about whether those AI projects pencil out when you add the real cost of risk.
The Implication
Watch for AI deployment to slow at mid-sized companies over the next 18 months. Not because the technology stopped working, but because the total cost of ownership just doubled when you add unsubsidized insurance. The companies that will keep moving fast are either big enough to self-insure (Google, Microsoft, Meta) or small enough that they're judgment-proof anyway (most startups).
The smart play for anyone building in this space: make insurance costs transparent in your pitch. If you're selling AI tools to enterprises, tell them what it actually costs to be covered. If you're deploying AI internally, run the numbers with your risk team before your next renewal. The era of free-riding on generic commercial policies just ended.