Reading time: 5 minutes
Listen time: 11 minutes
​

Good morning.

Two forces are reshaping litigation at once. Claims are getting more polished and more plentiful. The proof behind them is not keeping up.

Generative AI can draft a flawless-looking brief in minutes. It cannot manufacture the evidence that brief needs. That gap is where this week's intelligence lives.

Here's what's in front of you:

• Courts have now caught AI-invented citations in more than 1,700 filings (as of July 2026), and one judge has described the pace as rapidly escalating. A year ago the number was a fraction of that.
• One firm's briefs carried 15 fake cases and 8 invented quotes. The court didn't just fine the lawyers $110,000. It threw the client's whole case out with prejudice.
• You were told a federal rule for AI evidence was coming in 2026. The rulemakers just sent it back for a rewrite. Something has to fill the gap until it lands.

Five minutes. Let's get into it.

Prefer the audio this week? Press play here: Listen to this week's Brief

The cost of a lawsuit is climbing from both ends

Summary

Litigation is getting easier to file and costlier to fight at the same time.

Start with volume. Courts have flagged fabricated AI citations in more than 1,700 filings as of July 2026, up sharply over the past year (Damien Charlotin's tracker). One Oregon judge described the pace as rapidly escalating. Generative AI compresses the time between incident and complaint. A plaintiff firm can produce a complete, professional-looking brief in hours. Some of those briefs cite cases that do not exist. The filing count is rising regardless. For carriers and underwriting teams, that is a frequency driver, not a curiosity.

Now add severity. U.S. tort system costs reached an estimated $529 billion in 2022 and grew at roughly 7 percent a year over the prior decade, outpacing both inflation and GDP (U.S. Chamber Institute for Legal Reform / Brattle Group, 2024). In 2024, 135 verdicts exceeded $10 million against corporate defendants, totaling $31.3 billion. The median large verdict climbed to $51 million (Marathon Strategies, 2025 edition reporting 2024 data). Carriers booked $15.8 billion in adverse prior-year development on casualty lines, the highest figure on record. Put differently, insurers discovered their reserves were $15.8 billion short of actual losses. The aggregate spans all casualty lines, but when carriers close a gap that size, premium increases reach professional liability and every other line in the market.

The two trends stack. More filings raise the cost of clearing your docket. Bigger awards raise the stakes on every file.

So what?

Defense spend per claim is rising. Filing counts are rising. The two pressures compound. Reserving teams that model a flat defense-cost trajectory are working from assumptions the docket has already passed.

Monday morning question for your risk committee: are we pricing for a world where weak claims cost less to file and strong claims cost more to beat? If the actuarial team hasn't modeled AI-driven filing volume as a distinct frequency input, that conversation belongs on the agenda this quarter.

One more question for the table. Does your own use of AI in claims, underwriting, or filings create exposure your D&O or E&O program does not cover? The NAIC Model Bulletin on AI governance and a growing list of state rules make this a live question, not a future one.

AI can write a better brief. It cannot write the evidence.

Summary

The sanctions wave is the proof.

In 2023, a New York court fined two lawyers $5,000 for a brief full of cases their chatbot invented (Mata v. Avianca, S.D.N.Y.). That looked like a one-off. It wasn't. By December 2025, a federal court in Oregon hit two lawyers with $110,000, the largest AI-hallucination penalty on record, and dismissed their client's case with prejudice. The briefs carried 15 fake citations and 8 invented quotes, and no one had run them through a basic citation check (Couvrette v. Wisnovsky, D. Or., Dec. 12, 2025). The trajectory held into 2026. The Sixth Circuit sanctioned two attorneys $15,000 each in March (Whiting v. City of Athens, No. 25-5424). Courts logged at least $145,000 in AI-citation sanctions in Q1 2026 alone (ComplexDiscovery).

Researcher Viktor Wang describes this as "synthetic credibility." AI refines the packaging. The evidence underneath doesn't improve. A brief can read as airtight and cite nothing real.

The LION Lens

What happened — AI-hallucination sanctions climbed from $5,000 to $110,000 in roughly thirty months. The Oregon court dismissed the underlying case entirely, the first time a judge ended a client's lawsuit over fabricated AI citations (Couvrette v. Wisnovsky, D. Or., Dec. 12, 2025).

Why it matters — Courts are willing to end the case, not just fine the attorneys. At $110,000 the penalty registers on a firm's balance sheet. For the client, the case dismissed with prejudice is the real cost.

Practical implications — Clean opposing filings are not evidence of a strong case. They are a signal to verify what's underneath. AI can package what exists. It cannot manufacture what doesn't.

So what?

The defense approach that has worked on recent AI-heavy dockets refuses to argue on the plaintiff's terms. Reactive defense, answering each filing as it arrives, plays the game generative AI was built for.

Proactive defense reverses the dynamic. You choose what the record contains. You choose when the merits get tested. Wang's playbook maps three specific moves. Subpoena primary documents from neutral third parties so the record rests on sources the other side cannot rewrite. File corrective motions to strike unsupported claims before a narrative hardens. Signal trial readiness early to shift the settlement math.

For carriers, this translates to a claims-handling standard worth writing down. Ask panel counsel three questions. How do they verify opposing citations? When do they subpoena third parties? How early do they signal trial? If the answers are vague, that is your exposure.

For carrier buyers specifically, there is a second-order effect worth tracking. AI-polished plaintiff filings accelerate bad-faith allegations and extra-contractual exposure. When a complaint reads as airtight, a slow or clumsy claims response looks worse to a jury. That risk lands in your E&O and ICPL towers.

The LION POV

Here's how we're advising clients to handle litigation in an AI-heavy environment:

• Force the record through third-party subpoenas. Pull primary documents from neutral sources — insurers, contractors, banks — so the case stands on files no one can rewrite after the fact.
• File corrective motions early. Move to compel, strike unsupported filings, and fix admissions before a tidy narrative hardens into assumed fact.
• Signal trial readiness. A credible willingness to try the case shifts settlement math in your favor and calls the bluff on volume-based pressure.

Point your own AI at diagnosis, not narrative. Map the record. Flag gaps in the other side's filings. Stress-test their story against the documents. The winning move is making the claim prove itself.

These are starting points for your conversation with litigation counsel, not substitutes for case-specific strategy.

The federal fix for AI evidence just got shelved

Summary

Washington was expected to hand courts a rule for AI-generated evidence this year. It didn't.

Proposed Federal Rule of Evidence 707 would have put AI output offered without a human expert to the same test courts apply to expert testimony under Rule 702. The party offering it would need to show the AI used good data, sound methods, and applied them properly. Public comment closed February 16, 2026.

Then the rulemakers pumped the brakes. In its May 17, 2026 report, the federal Advisory Committee on Evidence Rules wrote that it "does not recommend action on the proposed Rule 707 at this time." It revised the draft and sent it, along with the problem of deepfakes, back for more study. This was not gridlock. In the same report, the committee gave final approval to a separate, unrelated rule change. It chose to wait on AI specifically.

A federal standard for AI evidence is now unlikely before 2027.

So what?

Do not plan your 2026 litigation strategy around a rule that is not arriving this year. Based on the committee's May 2026 report, a federal standard is unlikely before 2027 at the earliest. Until FRE 707 comes back, the backstop is the test you already have under Rule 702.

States and courts are not waiting. Several Texas federal district courts now require lawyers to certify that AI-generated content in their filings has been checked (e.g., S.D. Tex. General Order, 2025), and other jurisdictions have adopted or proposed similar rules. If you operate in more than one state, the question is not whether a federal rule arrives but which court rules already apply.

When the other side offers an AI-built damage model or exhibit, treat it as a question about method today. Make them show the data and the process. The federal rule is paused. The standard it aimed to set is already yours to demand.

The Bottom Line

AI changed how litigation looks. It did not change what wins.

The sanctions climbed from $5,000 to $110,000 in two years. Courts caught fabricated citations in more than 1,700 filings. The federal rule designed to address AI evidence went back for a rewrite in May. Through all of it, the burden of proof stayed where it was.

A claim still has to be proven with real evidence. The carriers that have achieved favorable outcomes on AI-heavy dockets are the ones forcing that proof, not the ones matching volume with volume.

What this means for your renewal. We are seeing early movement in the D&O and E&O markets in response to these trends. Some carriers have begun introducing AI-related endorsements or questionnaire requirements. Underwriters are asking about panel counsel verification protocols and internal AI governance. If your program renews in the next two quarters, raise these questions with your broker before the market raises them with you.

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/110-000-for-fifteen-fake-cases

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

Stay Covered Out There Y'all,

FLIP

Managing Partner
LION Specialty

P.S. Want to pressure-test how your panel counsel handles AI-era litigation? Our D&O Contract Vigilance Blueprint walks through the checkpoints, from citation verification to third-party subpoena timing. Reply "Blueprint" and we'll send it over.

P.P.S. Nothing in this briefing constitutes legal, coverage, or claims-handling advice. It is market intelligence designed to help you ask better questions of your advisors and make sharper decisions at your next insurance renewal. Consult your own counsel before acting on any recommendation.

If you didn't enjoy the email you can unsubscribe here.