Research · Benchmark Report 2025

The cost of one sentence.

Real, documented cases where a single line lost the lawsuit, the settlement, or the fine, defamation, antitrust, securities disclosure, product-liability admissions, run through VerbaPulse. Thirteen from the US, UK, and EU. The companies are anonymized, but the conduct, the figures, and the language are real, billions in combined cost, each traced to a line someone wrote in seconds.

Get the full report
Enter your details and we'll email you the full report: all 13 cases with the figures and what a pre-send check catches, straight to your inbox.
We'll send the report and the occasional research update. No spam, unsubscribe anytime.

On its way.

Thanks. We'll email the full Benchmark Report 2025 to shortly.

13
Real cases, every outcome documented
US · UK · EU
Three jurisdictions
2020–Q2 2026
Outcome window
Billions
In fines, settlements, and verdicts

A note on identities. Every case here is a real, documented legal or regulatory outcome. Company identities have been anonymized, and the names of any individuals have been removed or changed. The conduct, the figures, and the language patterns are unchanged.

Why we built this

A benchmark only means something when the inputs are real

Most tools that deal with communication risk arrive after the message is already gone: e-discovery, monitoring, the legal review that happens once a complaint lands. By then the message has been sent, stored, and is sitting in an inbox waiting to be read back in a courtroom.

So we ran the honest test. We took the actual language from cases that companies or their senior decision-makers lost, and put it through VerbaPulse at the one moment that still changes the outcome: before send. We did not write our own lookalike text. The identities are anonymized; the conduct, the figures, and the wording patterns are unchanged. This is the risk your organization carries today, in language your teams already use.

The cases

One line. The whole case turned on it.

Each row is a real outcome. The middle column is the kind of phrase that did the damage, the sort of language a pre-send check reads before it leaves the draft.

CaseChannelThe line that did itOutcome
A global aircraft manufacturer
US, 2021
Internal messages "I basically lied to the regulators (unknowingly)." USD 2.5B
DPA
A US cable-news network
US, 2023
Internal email A host, privately: "She is lying. I caught her." USD 787.5M
settlement
Two EU food-delivery platforms
EU, 2025
Inter-company email Proposed to "find some kind of non-solicitation agreement." EUR 329M
EU's first labour cartel
A video-game publisher
US, 2023
Internal messages Harassment-culture chat cited in the state complaint. USD 54M
settlement
A celebrity-founded sports company
US, 2025
Company statements That his age had "resulted in dementia." USD 50M
defamation verdict
A US college (vs a local bakery)
US, 2022
Written flyer "This is a RACIST establishment with a LONG ACCOUNT of RACIAL PROFILING." USD 36.59M
paid in total
A pharmaceutical company (France)
FR, 2022
Sales-force messages The generic "does not have the same composition" (overstating switch risk). EUR 25M
upheld on appeal
A group of UK broadcasters
UK, 2025
Email / messaging "...just want to be aligned and benchmark the rates." GBP 4.2M
UK's first labour cartel
A dominant ad-tech platform
US, 2025
Internal email "...as if a major bank owned the stock exchange." Monopoly
finding
A sports-betting company
US, 2024
Executive social post Posted material growth figures before results were released. USD 200K
SEC, Reg FD

The full report adds three more cases (a pharmaceutical company, a crypto exchange, and a blood-testing startup), each with the figures and what a pre-send check catches.

Anatomy of a flagged message

What the check returns on real text

This is actual VerbaPulse output on the internal messages from the aircraft-manufacturer case, not an illustration. The product reads the draft, marks the exposure with a severity and a plain reason, and recommends the action.

VerbaPulse scan · internal messages, aircraft-manufacturer case
Critical · Legal
"I basically lied to the regulators (unknowingly)."
Admission of lying to regulators.
Recommended: no compliant version preserves the intent, best removed. [Remove sentence]
High · Reputational
"This airplane is designed by clowns, who in turn are supervised by monkeys."
Disparaging language about the design team.
Recommended: no compliant version preserves the intent, best removed. [Remove sentence]
What the pattern shows

The same shape, across every category

Defamation, antitrust and labour-market collusion, securities disclosure, product-liability admissions, workplace conduct. Different laws, one through-line: an ordinary message, written in seconds, that became the centerpiece of the case.

Transparency

What we left out, and why

A benchmark you can trust line by line means showing the cases we excluded as clearly as the ones we kept.

FAQ

Common questions

What is the VerbaPulse Benchmark Report?
It studies thirteen real, documented cases from the US, UK, and EU in which a company or a senior executive lost a lawsuit, paid a settlement, or was fined, and the trigger was a specific piece of writing. Company identities are anonymized and individuals' names removed or changed; we show what a pre-send check flags on that kind of language.
Are the cases real?
Yes. Every case is a real, documented legal or regulatory outcome. We have anonymized the companies and removed or changed the names of any individuals, but the conduct, the figures, and the language patterns are unchanged.
Why are the companies anonymized?
These are real organizations' real mistakes. Even where the records are public, naming them to sell a product is not how we want to operate. Anonymizing keeps the lesson and the cost intact while leaving the parties their dignity. A fully sourced version is available to evaluating teams on request.
Does VerbaPulse claim it would have prevented these outcomes?
No. VerbaPulse flags the risky phrase at the moment before send and offers a safer line, or says plainly when the safest move is to remove it. Whether the story ends differently depends on the writer's choice. The benchmark shows where a check at the point of writing would have surfaced the exposure while the wording could still change.
How were the cases selected?
A case is included only if it is a documented legal or regulatory loss, the trigger was phrase-level content risk VerbaPulse addresses, the channel is one VerbaPulse can operate on, and the outcome falls between 2020 and Q2 2026.
Can I see VerbaPulse run on my own text?
Yes. Book a walkthrough and we will scan the kind of messages your team actually sends. The check runs in the browser inside Gmail, Outlook, and LinkedIn, and no message content is stored.
Sourcing

Drawn from the public record

Every case is a documented court judgment or regulator decision from 2020 to Q2 2026. We have removed the direct citations here to keep the parties anonymized. A fully sourced version is available to evaluating teams on request.

See it on the messages your team actually sends

Book a demo →