AI Is Supposed to Make Insurance Cheaper. For Liability Insurance, the Data Says the Opposite

What is actually driving the liability spike?

The headline figure from the market briefing is the one to sit with: liability pricing climbing 32.91% in roughly six months, between October 2025 and April 2026. Numbers like that are not the slow grind of a hardening market. They mark an inflection point, the moment a trend stops being gradual and starts compounding.

The usual explanations, inflation in repair and medical costs, social inflation, larger jury awards, are all still in play. But the episode's argument is that a newer force is now layered on top: artificial intelligence is lowering the cost and friction of litigation itself. When the plaintiffs' bar gets the same tools the carriers are celebrating, the supply of lawsuits goes up, and in liability lines, the volume and severity of claims is exactly what sets price.

A crucial point underlies all of this: you do not have to win a lawsuit to raise premiums. Even claims that are eventually dismissed carry defense costs, consume adjuster time, and add settlement pressure. Carriers price for the full distribution of what might come, not just for the cases that result in a verdict. So a flood of marginal, cheaply produced claims moves pricing even if most of them never reach a courtroom.

Channel 1: AI as a lawsuit enablement tool

The first channel is the one that changes claim volume across the whole market. AI is increasingly being used to find plaintiffs before they even know they have a case. Instead of waiting for an injured party to seek out a lawyer, the model scans data, identifies people whose circumstances fit a viable claim, and surfaces them as potential litigants.

That inverts the traditional economics of litigation. Historically, the friction of finding claimants, vetting them, and assembling a case acted as a natural brake on volume. Remove that friction with automation and the brake comes off. More identified claimants means more filings, more filings means more claims hitting policies, and more claims means upward pressure on liability insurance pricing for everyone in the pool, not just the defendant in any single suit.

Channel 2: AI as a fraud multiplier

The second channel is fraud, and the figure the episode cites is striking: 98% of carriers say AI is fueling a rise in fraud. Generative tools make it cheap and fast to produce convincing fabricated documentation, and to submit claims at machine scale. Bot-submitted claims and synthetic supporting paperwork can be generated in volume that manual review was never designed to absorb.

The cost of that does not vanish inside the carrier. It gets priced into renewals. When the expected fraud load on a book rises, the loss assumptions baked into pricing rise with it, and honest operators end up subsidizing the fraud the same way they subsidize every other modeled cost. This is the quiet version of the AI premium effect: not a dramatic courtroom event, just a steadily higher loss assumption flowing through to the number on your renewal.

Channel 3: AI as a liability generator

The third channel is the most direct: AI is not only a means of producing claims, it is becoming a subject of them. The episode points to a wave of AI-related class action filings hitting insurers directly, with 12 in the first half of 2025 alone, already exceeding the full-year 2024 total. A category of litigation that barely existed is now compounding fast.

These are claims arising from how AI systems are built and used, and they land on the liability and related coverage lines that commercial policies are built around. As the body of AI-specific case law grows, so does the uncertainty carriers have to price for, and uncertainty in an underwriter's model is just another word for a higher rate. This channel is also the one most likely to keep accelerating, because it feeds on the same AI adoption curve that is sweeping through every industry.

Where does this show up in the data, and why can't you escape it?

The reason these three channels matter to an individual operator rather than staying an abstract market trend comes down to structure. In a market with many competing carriers, rising costs in one place can be arbitraged away, because a buyer can move to a carrier that has not repriced yet. In a concentrated market, that exit does not exist.

New York is the example the episode uses. When only a handful of carriers are actively writing a given liability risk, and all of them face the same rising claim and fraud and class-action pressure, there is no competitor sitting on the sidelines to undercut them. The rising cost has nowhere to go but into the premium, and the buyer absorbs it in full. This is why two operators facing the same national trend can experience it very differently depending on how many carriers actually compete for their specific risk, and it is exactly the kind of thing that is invisible until you can see the carrier landscape behind your own placement.

What should an operator do before the next renewal?

The episode closes on action rather than alarm. Working from its framing, three moves follow directly for anyone who owns, operates, brokers, or lends against commercial real estate. First, find out where your liability pricing actually sits relative to the market, rather than accepting "this is market" at face value, because a 32.91% move is precisely when a reference point matters most. Second, understand the carrier concentration behind your specific risk: how many carriers genuinely compete for it, since that determines whether you have any leverage at all. Third, treat this as pre-renewal work, not renewal-day work, so you walk into the conversation already knowing your position instead of reacting to a number you cannot evaluate.

The common thread is visibility. None of these channels are things an individual operator can switch off, but the buyers who navigate this best are the ones who can see where their costs sit, who is behind their coverage, and where the leverage actually is, before the renewal arrives.