Reading time: 8 minutes
Listening time: 6 minutes
​

Your Wednesday Intelligence State of Play: US Bad Faith

​
​
​A Massachusetts judge turned a $26.6 million verdict into a $90,971,612 bad faith judgment.

The policy limit was irrelevant. In the same 12 months, three states made bad faith harder to prove, a federal court let plaintiffs argue that no human ever reviewed an AI-driven denial, and a Nevada ruling cracked the math on every layered D&O and cyber tower in the country.

Three reasons to read this edition.

1. Claims that are impossible to lose in some states are costing carriers $91 million in others. This edition maps which states moved and the claims-handling mistake that triggered the multiplier.
   ​
2. Plaintiffs stopped asking if the denial was fair. The question is whether a human looked at the file. The case law and the four documentation elements that will be the first discovery requests are inside.
   ​
3. Excess carriers can sue the primary for bad faith. The tower math shift and the communication protocol that would have prevented it are laid out below.

Prefer to listen? Check out the audio version.

The country just split in two on what bad faith even means.

The reform wave that started in Florida in 2022 and the expansion push building in the Northeast hit the same 12-month window. The result is not a slow drift. It is acceleration in opposite directions.

On the reform side, Florida created a 90-day safe harbor.

If a carrier tenders the lesser of policy limits or the demanded amount within 90 days of receiving notice with supporting evidence, bad faith shall not lie. Mere negligence, by itself, no longer establishes bad faith, and courts are now instructed to consider bad faith primarily as a damages question rather than an independent basis for liability. Georgia followed in April 2025 with a sweeping tort reform package that included tightened standards for time-limited demands, restricted duplicative attorney-fee recovery, and required disclosure of third-party litigation funding. Louisiana enacted its own reforms in August 2025. Three states, same direction, all within 14 months.

Then there is Massachusetts.

On December 1, 2025, a Suffolk County judge ordered three Liberty Mutual entities to pay $90,971,612 in bad faith damages. The underlying personal injury verdict was $26.6 million for a 2014 construction accident. Liberty clung to a debunked defense theory and failed to settle after liability became reasonably clear. The court found that Liberty's conduct met the statutory standard for willful or knowing violations, characterizing it as willful blindness.

That finding triggered the mandatory doubling provision under Massachusetts's consumer protection statute (G.L. c. 93A), applied to the full underlying judgment. Not the limit...the verdict! That is how $26.6 million becomes $91 million. The judgment is reportedly still accruing interest.

The outcomes are not all running one direction.

In 2026, a federal jury in Florida returned a complete defense verdict for Kinsale after more than a decade of bad faith litigation on a wrongful death failure-to-settle claim. The difference between a $91 million judgment and a complete defense verdict came down to the same variable: how the carrier documented its claims-handling decisions.

New York has S166 in committee.

It's a bill that would create the state's first modern unfair claim settlement practices statute with a private right of action for bad faith, including compensatory and punitive damages plus attorney fees. The bill was re-referred to the Insurance Committee in January 2026, and whether that signals momentum or procedural delay remains an open question.

South Carolina introduced H.B. 4733, which would create a rebuttable presumption of bad faith in certain settlement contexts if enacted.

In Texas, the exposure extends beyond insurers and MGAs are no longer immune.

In CiCi Enterprises v. HSB Specialty Insurance Company (N.D. Tex. 2026), a federal court allowed bad faith claims to proceed against both the carrier and its MGA, At-Bay Insurance Services. The claims arose under Texas Insurance Code Chapter 541, which prohibits unfair settlement practices by insurers and their agents. At-Bay had drafted the policy form and administered the claim under delegated authority through its in-house claims team.

The bad faith allegations centered on coverage communications handled by the MGA under delegated authority, with limited indication the carrier directly approved the language. Where an MGA handles claims under binding authority, bad faith exposure can originate in one organization and land on another.

The mistake in Massachusetts was a claims-handling failure.

Liberty rode a defense theory that had already collapsed and the multiplier turned that decision into a catastrophe. The mistake in Texas was structural. An MGA's claims operation generated bad faith exposure the carrier absorbed without visibility into the communications that created it. In Massachusetts and every state with a multiplier regime, bad faith damages are calculated against the verdict.

The carrier is risking whatever number a jury writes down, doubled. In every MGA program with delegated claims authority, the carrier is also exposed to whatever its MGA communicates to the policyholder during the claim. Both paths lead back to the same place: the claim file.

If you don't know which multiplier and safe-harbor regimes apply to your book, reach out and we'll map your jurisdictional exposure across your full program.

The split is only the first axis. The next force ignores state lines entirely.

Plaintiffs stopped arguing your AI got the denial wrong. Now they argue no human ever checked it.

The regulatory wall around AI in claims handling is closing now.

Not next cycle. This year. Twelve months ago, a carrier could deploy AI across its claims operation without documenting the algorithm's role in any individual coverage determination. That window is shutting state by state, and the plaintiffs' bar found the theory to exploit it before the regulations even finish landing.

An estimated 82% of major insurers use AI somewhere in the claims process.

Not all of it carries equal bad faith risk. An algorithm that flags potential fraud for human review operates in a different risk category than a model that recommends a coverage determination or calculates a payout amount. The emerging bad faith theories target the second category: AI that participates in the decision, not AI that supports it.

In Estate of Lokken v. UnitedHealth Group (D. Minn. 2025), a federal court allowed breach of the implied covenant of good faith and fair dealing to survive.

The surviving claim rested heavily on one core allegation: the insurer never disclosed its use of AI in rendering claims decisions. No human being had been shown to have independently evaluated the file, and that combination kept the case alive. Lokken arose in health insurance, where AI-driven claims processing is most advanced, but the underlying theories are portable to P&C and financial institution lines.

The question in bad faith litigation moved from "did the insurer ignore the evidence" to "did a qualified human ever engage with this specific claim."

Regulators are accelerating that shift. California's SB 1120, effective January 2025, prohibits health insurers from denying coverage based solely on an algorithm. Florida's HB 527 would mandate independent human review of every AI-assisted claim denial and is advancing through committee. Colorado's AI Act enforcement begins June 2026, requiring mandatory algorithmic inventory and bias testing.

The NAIC AI Model Bulletin, now adopted in 24 states plus D.C., identifies four documentation elements carriers should treat as the emerging baseline:

• the AI's specific role in each claim decision
• the identity and credentials of the human who reviewed the output
• the override rate for AI-assisted denials
• the disclosure language provided to claimants about AI involvement

Those four elements could be the first discovery requests in any AI-related bad faith action.

The mistake: treating the AI output as the decision rather than as input a human professional must own.

Ratifying an AI recommendation is not the same as reviewing the claim.

Courts will measure whether a qualified professional independently evaluated the specific file and documented that evaluation, not whether someone approved a recommendation without examining it. The standard applies equally to carrier claims departments and to MGAs running AI-driven claims operations under delegated authority.

If you cannot document the human review, you cannot defend the process.

For any organization running AI in claims, this belongs on the risk committee agenda now: does your board know which claims workflows use AI, and has legal signed off on the documentation and disclosure framework?

The AI question changes how claims get judged. The next question changes the structures you actually buy.

Your D&O insurance tower now has a fault line running between its own insurers.

D&O and cyber programs are built as towers. A primary layer, one or more excess layers, sometimes a buffer. The whole structure assumes the layers pull in the same direction when a claim hits.

A 2026 ruling from the Nevada Supreme Court broke that assumption.

In North River Insurance Co. v. James River Insurance Co. (Nev. 2026, en banc), an apartment complex insured under a $1 million primary policy and a $10 million excess policy faced a fatal shooting claim. The deceased's estate made two settlement demands within the primary limits. The primary insurer declined both. The claim ultimately settled for $5 million. The excess insurer then sued the primary for equitable subrogation — the legal right to step into the insured's position and bring the same claim — arguing the primary's failure to settle within its own limits breached the duty of good faith.

The Nevada Supreme Court agreed.

An excess insurer can now stand in the shoes of the insured and bring a bad faith claim directly against the primary carrier. The underlying lawsuit settled within the combined primary and excess limits. It did not matter. The excess carrier paid more than it should have because the primary refused two reasonable demands, and the court held that was enough.

The mistake is assuming that a settlement within combined tower limits protects the primary.

In Nevada, it does not. Whether other states adopt this framework depends on their equitable subrogation standards, but the reasoning is portable and plaintiffs' counsel in layered programs now have a roadmap.

Consider what this looks like in a real D&O or cyber tower.

A claim comes in. The primary carrier controls the defense and the settlement authority. The primary declines a within-limits demand because its own exposure feels manageable. The case settles higher. The excess carrier absorbs the overage and then turns around and sues the primary for the difference.

Before North River, that cause of action had not been recognized in Nevada. Now it has.

The practical starting point:

• Establish a documented communication protocol with excess carriers at the inception of any claim that could implicate the tower
• Notify excess layers when a within-limits demand is received
• Document the primary's settlement authority and the basis for any declination
• Confirm in writing that excess carriers have visibility into the claim's trajectory

North River turned on two declined demands the excess carrier learned about too late.

Your primary carrier's next settlement call could become your excess carrier's next complaint.

We review D&O and cyber towers for exactly this kind of structural exposure. If you want to know which counterparties in your current program could create the same problem North River did, let's walk through your tower.

In Case You Missed It!

This week’s Wednesday Intelligence maps the 2026 bad faith landscape, including the AI theory plaintiffs are now using to bypass your claims process entirely.

If that section caught your attention, our Six-Line Silent AI Audit series covers the other side of the same coin: the coverage gaps AI tools are quietly creating inside your own program. All three parts are live.

Part 1: D&O and EPLI, where “wrongful act” definitions assume a human decided. Part 2: E&O and Cyber, where the liability boundary for AI-assisted advice is unsettled and deepfake wire fraud falls between coverage sections. Part 3: the full audit framework across Fiduciary and Crime/FI Bond, plus the governance docs leading FI writers are asking for at renewal.

Read Part 1 here, or listen to the audio version here.
Read Part 2 here, or listen to the audio version here.
Read Part 3 here, or listen to the audio version here.

Thank you for reading today’s edition!

Want to share this edition via text, email or social media? Simply copy-and-paste the link below:

​https://lionspecialty.kit.com/posts/an-excess-insurer-can-now-stand-in-the-shoes-of-the-insured-and-bring-a-bad-faith-claim-directly-against-the-primary-carrier​

And if this briefing was forwarded to you, subscribe directly here.

Stay Covered Out There Y’all,
​FLIP
​Founder and Managing Partner
LION Specialty

P.S.S. Nothing in this briefing constitutes legal advice. These are the opinions of the founder. It’s market intelligence designed to help you ask better questions of your advisors, your carriers, and your broker at your next renewal. If you loved it, consider telling your colleagues to subscribe.