(610) 804-8021 info@centroid-ai.com

Counterfeiting in Pharma – The Big Asymmetry

CentroidAI
A FyndEm Shield Perspective · 2026
An industry perspective on pharmaceutical anti-counterfeiting

Four Asymmetries,
One Advantage.

Four structural asymmetries separate the counterfeiter from the manufacturer — in speed, in posture, in tempo, and in visibility. Between them they explain the spend problem, the ROI problem, the iceberg problem and the whack-a-mole problem. And there is one advantage the manufacturer already holds, which almost no one is using.
01

The state of play

Counterfeiting is a fast-growing, dangerous risk to the pharmaceutical industry, endangering our patients’ safety.

7,152
Pharmaceutical crime incidents recorded in 2025 — the highest annual total ever recorded
Pharmaceutical Security Institute, CY2025
142
Countries impacted by pharmaceutical crime
Pharmaceutical Security Institute, CY2025
93
Different medicines discovered in a single incident
Pharmaceutical Security Institute, CY2025
+11%
Year-over-year growth in recorded incidents
Pharmaceutical Security Institute, CY2025
2,857
Different medicines involved across those incidents
Pharmaceutical Security Institute, CY2025
+29% / +18%
Growth in recorded incidents — Europe, and Asia-Pacific, respectively
Pharmaceutical Security Institute, CY2025 — geographic distribution

For wider context, the OECD and EUIPO put global trade in counterfeit goods at an estimated $467 billion, with medicines among the fast-growing, higher-risk categories — a figure that spans all product categories and is based on 2021 trade data.

And these are the incidents that were detected and reported. Many brand-protection leaders share the same uncomfortable suspicion: what is seen is a fraction of what exists. That suspicion is the first clue to the real problem.

02

The cost is real — and mostly unmeasured

Most counterfeit cases never get a public price tag. The few that do make the stakes unmistakable — measured in dollars, in trust, and in lives.

By the dollars

A $230–250M counterfeit HIV-drug ring

A criminal network tampered with and counterfeited genuine HIV medications — including widely used antiretrovirals — and fed an estimated 80,000+ altered bottles back into the legitimate U.S. pharmacy supply chain over roughly two years. The ringleader was sentenced to 15 years and ordered to forfeit $238.8M.

Reported by Reuters, CNBC; U.S. DOJ proceedings · scheme ~2019–2021, sentenced 2023
By lives lost

78+ dead from a falsified ingredient

Industrial solvent diethylene glycol was falsely labeled as pharmaceutical glycerin and used to make cough and cold syrup distributed to unsuspecting patients. At least 78 deaths were documented, with estimates ranging higher, and lasting harm to survivors.

Documented by The Pew Charitable Trusts · Panama, 2006
By scale of harm

Counterfeit pills, tens of thousands of deaths

Criminal networks press fake prescription pills that look identical to legitimate medicines, many sold through illicit online pharmacies. In 2022 an estimated six in ten seized counterfeit pills carried a potentially lethal dose of fentanyl. Overdose deaths with evidence of counterfeit pill use more than doubled nationally between 2019 and 2021, and more than tripled in western states — to 14.7%.

U.S. CDC (MMWR, September 2023) and DEA · data through 2022

These three were documented because someone — a prosecutor, a regulator, a public-health agency — finally counted. The overwhelming majority are never counted at all. Which is exactly why the true magnitude stays hidden, and why the industry keeps under-reacting to it.

03

The asymmetry

Step back from the individual incidents and a structural picture emerges. The counterfeiter and the manufacturer are not playing the same game, and they are not playing by the same rules. Four asymmetries separate them — and not one of them is accidental. Each is a direct consequence of the fact that one side is regulated and the other is not.

The counterfeiter
Plays offense. Unconstrained.
  • Speed. Re-tools in weeks. No qualification, no filing, no change control — aided by 3D design, on-demand printing and AI
  • Posture. Chooses the time, the place, the product and the market. Attacks where the defense is thinnest
  • Tempo. Acts first, every time. Sets the pace and forces the response
  • Visibility. Conceals deliberately. Knows exactly what it is doing, and where it will move next
The manufacturer
Plays defense. Constrained by design.
  • Speed. Changes in months to years. Every feature specified, qualified, validated, documented and filed before it ships
  • Posture. Must be ready everywhere, on every product, in every market, at once
  • Tempo. Responds to what has already happened — and every new defense begins being copied the day it ships
  • Visibility. Sees only what surfaced: a complaint, a seizure, a return. Guesses where the next strike lands
The counterfeiter moves at software speed.
The manufacturer moves at validation speed.
This is the asymmetry. Every other problem in pharmaceutical anti-counterfeiting is a symptom of it.

They can see you. You cannot see them.
With FyndEm Shield, you can now see them — before harm is done.

CentroidAI · Four Asymmetries, One Advantage · 1–2 / 6

04

The four symptoms

Most brand-protection leaders recognize these. What is harder to see is that all four trace back to the same root cause.

The blind-spend problem

Millions of dollars flow into overt and covert packaging features. Yet few manufacturers can confidently answer the simplest question: which of these features actually reduces counterfeiting?

↳ a symptom of investing without measured effect

The ROI problem

Investment committees do real diligence — but they decide largely without the one input that would make the call rigorous: evidence of what actually deters, across your own products and markets. Effect is hard to measure, so return is hard to prove.

↳ a symptom of having no measured effect to reason from

The iceberg problem

Confirmed incidents are the tip. The true volume below the surface is uncertain — so exposure tends to be under-estimated, and the business can under-react.

↳ a symptom of seeing fragments, rarely the whole

The whack-a-mole problem

The goal is to remove the operation. The reality is removing its artifacts — a marketplace listing, a business registration, a website domain — while the entity behind them remains unresolved and simply spins up new ones.

↳ a symptom of acting on artifacts, not the entity

Four problems. One cause. The manufacturer is fighting reactively, in fragments, an adversary that acts first and hides where it will act next.

05

Why the current model keeps losing ground

When counterfeiting rises, the instinct is to add another packaging feature. It feels like action. But packaging is only one front — and on every front, a static, reactive, after-the-fact response loses ground to an adversary that re-tools faster than the defense can be qualified:

Too late
Defenses are added after a threat surfaces — with no lead time, the counterfeiter has already moved and sold.
Too blind
Packaging features are copyable, and their effect is rarely measured — so spend continues without evidence of what works.
Never closed
When action is taken, few measure whether it worked — so the same gaps and the same sellers recur.
CentroidAI · Four Asymmetries, One Advantage · 3–4 / 6

06

The shift a manufacturer can make

Closing the visibility asymmetry does not necessarily mean spending more. It means changing the shape of the response — from the left column to the right.

DimensionToday — defensivePossible — predictive
PostureReact after a counterfeit surfacesPredict the threat as it forms; act before it reaches a patient
VisibilityYou learn where they struckYou see where they are forming — while there is still time to act
ScopeOne brand, one product at a timeYour whole portfolio — and the patterns that run across it
Feature spendDiligent, but without measured effectInformed by which features measurably deter, with quantified ROI
ExposureCounted from confirmed incidentsEstimated honestly, including what lies below the surface
The threatA static event — a seizure, a complaintA pattern tracked as it develops and moves across geographies and products
TargetIndividual listings, registrations, domainsThe resolved entity behind them

Each row is a move from reacting in fragments toward predicting from your own record — each one helps flatten the asymmetry. And unlike a point-in-time assessment that decays as the adversary adapts, an intelligence approach learns and compounds over time.

07

How the manufacturer takes the offensive

Speed, posture and tempo are structural. They are not going to change, and they should not. Visibility is different — and there is an advantage the manufacturer already holds that turns it around completely. It runs entirely in the manufacturer’s direction, and almost no one is using it.

The new approach

The counterfeiter cannot see what it leaves behind. You can.

To reach a patient, a counterfeit has to move through your channel — or visibly around it. The counterfeiter can conceal who it is. What it cannot conceal is the shape of its movement: inside your channel it meets a record kept end to end at every handoff, and outside it, it has to advertise in order to sell.

And it has no telemetry on itself. It does not know what it is leaving behind, which corridors it has already made conspicuous, or how legible its own pattern has become. You have spent a decade, and a very large sum, building the record that holds all of it. Serialization was implemented to satisfy a regulation. It also happens to be the most complete map of your own supply chain that has ever existed.

This advantage runs in your direction — and using it does not require new data. It requires using the data you already own, before the counterfeit reaches the patient.

What inverting visibility actually looks like. FyndEm Shield reads the serialization and distribution record a manufacturer already holds, and turns four questions that are answered late into four that are answered early. Predict, prevent, and — the part that has never been possible — prove:

Where it is forming

Today a seizure, a complaint or a return tells you where a counterfeit already struck — after the product sold, and after a patient was exposed.

↳ the concentration, and the corridors it is moving through, surfaced as it forms — with lead time to act

How much is actually at stake

Today exposure is counted from confirmed incidents, so the figure is always an undercount — and the business under-reacts to it.

↳ magnitude estimated from your own configuration and your own data, not from industry averages

Whether your defenses are working

Today a packaging feature is specified, qualified and shipped — and its deterrent effect is assumed rather than measured.

↳ measured deterrence, per feature and per combination, with a review date

Who is actually behind it

Today you act on artifacts — a listing, a registration, a domain — while the entity that created them spins up replacements.

↳ the resolved entity behind the artifacts, in a case package built to support enforcement

And this is not another major program. You do not have to instrument, purchase or build anything. The foundation already exists — the serialization and traceability record the industry has spent the past decade, and an enormous amount of capital, implementing. It is already captured, already validated, already inside your systems. Shield reads that record against the public market signals a counterfeit operation has to leave behind in order to sell, and it is the combination of the two that produces lead time. What has been missing is the decision to read it.

Where this leads

We built the answer to the asymmetry.

CentroidAI’s FyndEm Shield is a predictive intelligence platform for pharmaceutical anti-counterfeiting, built to run for a single manufacturer inside that manufacturer’s own environment. It is built to close exactly the gap this paper describes: prediction over reaction, the portfolio over the product, measured effectiveness over blind spend, the resolved entity over the listing — and the serialization record you already own, finally put to work. It runs read-only, inside your perimeter, and nothing is written back to your systems. How it does that is a conversation we would welcome.

Raj Subramanyam · Founder & CEO, CentroidAI  ·  Meet us at the HDA Traceability Seminar, Washington DC  ·  raj@centroid-ai.com
CentroidAI · Four Asymmetries, One Advantage · 5–6 / 6
About this paper. This is an industry perspective authored by CentroidAI. Figures cited are drawn from public sources — the Pharmaceutical Security Institute (CY2025 incident data, including its geographic distribution reporting), the OECD–EUIPO report Mapping Global Trade in Fakes 2025 (based on 2021 trade data), the U.S. CDC (MMWR, September 2023) and DEA, U.S. Department of Justice proceedings as reported by Reuters and CNBC, and The Pew Charitable Trusts — and are used for context, not as precise measures of any company’s exposure. The $467 billion figure refers to global trade in counterfeit goods across all categories, not medicines alone. The paper describes an industry-wide structural problem and the shift a single manufacturer can make to address it; it does not disclose proprietary methods, and it references no specific manufacturer. Where the scale of unobserved counterfeiting is discussed, it is explicitly framed as estimated and uncertain. Any figures shown in demonstration are illustrative and drawn from synthetic data.