Originally published on Medium.
Falsified Ozempic was caught inside the legal EU supply chain in 2023, which is the strongest case against the argument that printed numbers prove nothing. Look at what actually caught it.
Yesterday we argued that a batch number cannot prove what a product is, because it is ink and ink can be printed again. The obvious objection is that the pharmaceutical industry solved this years ago. Europe requires nearly every prescription pack to carry its own unique serial number, checked against a repository before it is handed to a patient. The United States requires a serial number on every package too. So the problem is already fixed, at least where it matters most.
That objection deserves a real answer, and it starts with the case that supports it.
On October 18, 2023, the European Medicines Agency alerted patients and health professionals to falsified Ozempic pens found at wholesalers in the EU and the UK. The pens carried, in the agency’s words, “batch numbers, 2D barcodes and unique serial numbers from genuine Ozempic packs.” And the system worked. “When the packs of the falsified Ozempic were scanned, the serial numbers were shown to be inactive,” which alerted operators to a potential falsification.
That is a documented case of per-pack serialization catching falsified medicine inside the legitimate supply chain. A batch number alone would not have caught it, because the batch number on those pens was genuine. We should say so plainly before saying anything else.
Now look at what did the catching.
The code proved nothing. The ledger entry did.
Every number on those pens was correct. The batch number was real. The serial number was real. The 2D barcode scanned cleanly. No printed number on them would have revealed anything wrong. The pens did look different from genuine ones, but that difference only helps someone who knows what a genuine pen looks like and is looking for it.
What revealed it was a state held somewhere else: the repository showed those serial numbers as inactive. The pens failed a lookup, not an inspection.
That distinction is the whole subject of this piece, because it tells you exactly when the mechanism works. It works when the database already knows. It does not work when the database does not.
What a lot number is, according to the people who define it
Before the database, a word on the number itself, because GS1 has settled this on the record.
The GS1 General Specifications define Application Identifier (10) as the batch or lot number, and state that it “associates an item with information the manufacturer considers relevant for traceability of the trade item to which the element string is applied.” It may be “a production lot number, a shift number, a machine number, a time, or an internal production code.” Then comes the note, in the standards change notice that introduced the current wording in April 2019: “The batch or lot number is not part of the unique identification of a trade item.”
The standards body does not claim a lot number identifies anything uniquely. It says the opposite. Regulators recall by it, distributors accept goods against it and consumers are told to read it off the jar anyway.
The regulators themselves keep confirming the gap in their own alerts. The World Health Organization’s alert of June 19, 2024 states that for one falsified batch, “batch number MP5E511 is genuine, but the product is falsified.” The US Food and Drug Administration’s alert on counterfeit Ozempic notes that lot PAR1229 is “also an authentic lot number,” and that patients who received genuine product from that lot through the manufacturer’s assistance programme may keep using it. Both agencies are explaining, in a safety notice, that the number cannot sort the real from the fake.
Billions of scans a year, eleven confirmed cases on the EMA’s books
So the serial number and its database are the better system. The question is how much it has found.
The European Commission’s report to the Parliament and the Council on trends in the falsification of medicinal products, published July 4, 2024, cites the EMA’s count for centrally authorised medicines: “11 confirmed cases of falsified medicines since 2019.” For scale, EMVO reports the European Medicines Verification System handling roughly 14 billion transactions a year across some 1,800 connected marketing-authorisation partners and more than 145,000 end users.
Two honest caveats on that number, both of which matter. The Commission does not say those eleven were detected by the verification system, and it says directly that “it is not possible at this stage to draw clear conclusions on the actual ability of the EMVS to effectively detect all falsified medicinal products throughout the supply chain.” And the figure does not reconcile with every other published count: German trade press, summarising the same EMA analysis, reported 67 confirmed falsifications in the legal EU supply chain between 2017 and 2022, 21 of them in the Netherlands and 13 in Germany, while noting that member states define a confirmed case differently. Different scopes, different definitions, no reconciliation available. What every version has in common is the order of magnitude: tens of cases, not thousands.
The Commission’s own commissioned study, completed in December 2023, found that of 16 national competent authorities that shared data, only two reported a single confirmed case each since 2019: Belgium in 2020 and Finland in 2019. EMA’s confirmed cases, which cover centrally authorised medicines, fell from 24 in 2017 to 3 in 2022, averaging about 30 packs per case. And the study is careful to offer the alternative reading, which is that the low count may reflect incomplete implementation rather than an absence of fakes.
The alerts are almost all noise, and everyone says so
The system does generate alerts in volume. The Commission reports that EMVO’s target is an alert rate below 0.05 percent, and that in seven countries the rate ran between 0.16 and 1.17 percent. EMVO names three main causes: errors in the data uploaded by the marketing authorisation holder, scanner issues, and double decommissioning. The Commission’s assessment is blunt: “Many automated alerts are not subsequently confirmed in the form of falsified medicinal products,” and “some IT systems developed at national level still trigger a level of false alerts that is too high.”
The cleanest published breakdown comes from Bulgaria. A peer-reviewed study of Bulgarian national data, published in May 2026, found that in the last quarter of 2024 alerts accounted for 0.062 percent of all scans, of which 73.57 percent were attributed to technical reasons and 26.43 percent to procedural issues. Neither category is falsification.
This is the operational reality the mechanism lives in. Pharmacy staff see alerts constantly, almost none of which mean anything, which is the environment in which a real one has to be noticed.
Where the ledger runs thin
Because the mechanism depends on database state rather than on the pack, it fails wherever the database is incomplete. The published record is specific about where that is.
Packs that are never checked out. The Commission reports, on EMVO data from September 2023, that 26 percent of medicines carrying safety features are not checked out of the repository system when dispensed. Those packs leave the chain with their serial numbers still active, and an active serial number is exactly what a cloner needs.
Aggregation. The Commission’s safety features Q&A, current to July 2026, states that “aggregation is not regulated in the Delegated Regulation” and is “only voluntary.” Nothing requires a case to carry a readable list of the packs inside it, so verifying a pallet means opening it.
Wholesale verification. The same Q&A describes wholesaler verification as a minimum requirement, and exempts packs shipped directly from the manufacturer, the marketing authorisation holder, or a designated wholesaler.
The United Kingdom. Great Britain left the EU system when the FMD ceased to apply there at the end of 2020, and since January 1, 2025, under the Windsor Framework, the FMD safety features no longer apply in Northern Ireland either. The MHRA’s guidance states that “there will be no serialisation in the UK,” though 2D barcodes and serial numbers remain permitted. A major European market has left the system entirely.
The United States, which never had the dispense check. US law requires a product identifier containing a serial number, but verification is required on suspect product, on saleable returns, and on request, not on every pack handed to a patient. And on August 6, 2026 the FDA extended an exemption for small business dispensers to November 27, 2027, under which transaction information may omit the package-level product identifier altogether.
The first scan wins
There is one more limit, and it is the one that matters most for anyone relying on a scan at the counter.
A cloned serial number that is still active passes. The system detects the clone only when the same identifier is presented a second time, and only if both packs are actually presented. Writing in a 2017 point-of-view article in the journal Medicine Access @ Point of Care, Roberto Frontini called this the scheme’s biggest weakness: criminals holding a list of codes “could put the UI on falsified packages and enter the drugs into the market,” and the first code decommissioned “will be recognized as authentic even if it was on the falsified package.”
Read that against the 26 percent figure. Whichever pack reaches a scanner first is the one the system calls genuine. In the 2023 case the cloned numbers had already been deactivated by the time the fakes were scanned. Nothing in the design guarantees that order.
The objections, briefly
So serialization does work. Sometimes, and better than a lot number. The Ozempic case is real and we are not arguing it away. The claim here is narrower and verifiable: it works by database state, not by anything checkable on the object, so its performance tracks how complete the database is.
Eleven cases might mean there is very little falsification in the legal chain. It might. The Commission says most falsification is detected in the illegal supply chain, and that is a genuine possibility the evidence does not exclude. Its own study offers the other reading too. Both are live, and neither is a reason to treat a scan as proof.
The alert noise is an implementation problem, not a design problem. Largely true, and it is being worked on. It is still the condition under which the check is actually performed today, seven years in.
A better code would fix it. Codes are printed. The pens in 2023 had perfect codes. What they lacked was any property of the physical pen that a counterfeiter could not reproduce along with the carton.
What each party should be able to prove
For a manufacturer, the test is whether you can distinguish your unit from a perfect copy of your unit when the database has no entry either way, which is the situation at every point before the first scan and in every market outside the system.
For a regulator, the test is whether a verification result means what the person reading it thinks it means. An active serial returns a pass for a genuine pack and for a clone of that pack, and only the sequence of presentation separates them.
For a wholesaler or pharmacy, the test is what you do with the alert you will almost certainly see this week, knowing that, in the one national breakdown published, roughly three quarters were technical, and that the real one looks the same.
For anyone designing the next scheme, whether that is a digital product passport or a track-and-trace rule, the test is whether you are building a register of numbers or a way to check an object. Europe built one of the most complete registers of pharmaceutical numbers in the world. It is a real achievement, and it tells you whether a number is active. It cannot tell you whether the thing in your hand is the thing the number was issued for.
Aeroz, Making Authenticity Checkable. Visit aeroz.io to learn more & get in contact with our team via [email protected].
© 2026 Aeroz. All rights reserved.