Originally published on Medium.
Singapore found a banned dye in two batches of a Korean face cream. The maker tested the same two batch numbers, found nothing, and said it had never shipped those numbers to Singapore.
On August 28, 2026, Singapore’s Health Sciences Authority published the results of its own laboratory testing on a Korean face cream. Three companies market the Medicube PDRN Pink Collagen Capsule Cream in Singapore, and HSA had sampled the product from all three. Its announcement is precise: “HSA’s testing detected minute amounts of Sudan IV in two samples of the product sold by Venus Beauty Pte. Ltd. (Batch nos: 2E122I.2E117I and 2E191G.2E193G).” Sudan IV is an industrial dye, prohibited in cosmetics. HSA had ordered those two batches recalled two days earlier, and cleared the batches submitted by the other two sellers.
The trail had started a month before, when Hong Kong’s Customs and Excise Department bought the cream during retail spot checks and warned the public on July 24 that it had tested positive for a banned Sudan dye.
Then, in the first week of September, the maker answered with numbers of its own. APR, the Korean company behind Medicube, said tests at certified institutions, including one accredited by HSA at the regulator’s request, found no dye in product carrying batch numbers 2E122I.2E117I and 2E191G.2E193G. It went further. It said it had never supplied or exported any product to Venus Beauty, and had no record of exporting those batch numbers to Singapore at all, according to the Korea Herald. Its own summary of the position, reported by Seoul Economic Daily, was that it is difficult to verify whether the item tested was genuine, and that it is unclear where the item came from or how it was distributed and stored.
So a national regulator and a listed manufacturer each tested batch 2E122I.2E117I and got opposite results. As far as we can find, neither has withdrawn, and six weeks later no public source establishes whose jars those were.
Both can be right, and that is the whole point. A batch number is not a property of the jar the way a fingerprint is a property of a hand. It is ink.
What a batch number actually is
A batch or lot number exists to let a manufacturer find a production run in its own records. If something goes wrong at the filling line on a Tuesday, the number is how the company works out which pallets to pull. It was designed as an internal index, and for that job it is excellent.
Everyone else has quietly repurposed it as an identity. Regulators recall by it. Distributors accept goods against it. Pharmacists check it. Consumers are told to read it off the base of the jar and compare it to a notice. None of those uses were in the original design, and all of them assume the same thing: that a given number appears on exactly one set of real objects.
Nothing enforces that assumption. The number is printed, which means it can be printed again, on anything.
Four ways a printed number fails
Enforcement records from the past two months show each failure mode separately, which is useful, because they need different answers.
The number does not exist. This is the crude version. In Kano, Nigeria, NAFDAC intercepted a counterfeit antimalarial during routine market surveillance and, in an alert released on August 19, said its registration number, “02 3457,” does not exist in the agency’s database, as Brand Spur reported. It was printed because genuine packs carry one. It was caught by a regulator checking its own database during market surveillance. NAFDAC’s public Green Book would let a patient look the number up, but only one who thought to.
The number is real, correctly formed, and on the wrong product. On September 23, the Thai FDA announced raids on four sites in Samut Sakhon and Bangkok producing counterfeit Allergin syrup, an antihistamine. Officers seized about 37,000 finished bottles, more than 150,000 label stickers, more than 850,000 caps and more than 400,000 empty bottles. The counterfeit bottles carried registration number 1A 271/62, which belongs to the genuine product registered to N&N Pharmaceutical Co., Ltd., along with lot number L2607061, a manufacturing date of 31.07.2026 and an expiry date of 30.07.2028. Everything a buyer could read was complete, plausible and internally consistent. Testing found no chlorpheniramine and no other active ingredient. The bottles contained flavoured, coloured syrup.
The number is real and its dates have been overprinted. Mexico’s health regulator COFEPRIS issued an alert on September 28 covering counterfeit Saizen, Bedoyecta, Buscapina Compositum and the wound spray Granulox. Rio Times reported that several of the fakes carry real lot numbers whose original expiry dates have passed, printed with new ones. Sanofi said Buscapina lots that expired in 2022 and 2024 now carry dates between December 2026 and December 2029. Granulox lot 16115143, originally expiring November 2019, later appeared marked December 2026. The same pattern shows up outside medicine: among the Norfolk seizures CBP listed on September 29, 1,176 rolls of carpet carried OEKO-TEX textile safety labels whose certification number had expired.
The number is real and now names two different things. This is the Medicube case, and it is the hardest of the four, because nobody has lied about the number. HSA tested an object carrying 2E122I.2E117I. APR tested an object carrying 2E122I.2E117I. Both reported accurately. The number stopped being a unique handle on a thing, and once that happens, a test result about one object says nothing reliable about the other.
The industry has already written this down
This is not a prediction. On October 9 the Korea JoongAng Daily published a survey of the trade in counterfeit Korean cosmetics, and its list of what the fakes now replicate reads: packaging, batch numbers, sealing stickers and pricing. An expert quoted in the piece says none of the tools is foolproof and that telling a fake from a genuine unit may require analysing the formula itself. The same reporting notes that production is divided up, with separate suppliers making the contents, the containers and the boxes, which is exactly the structure that makes a printed number cheap to reproduce at scale.
The more interesting detail in that piece is what the better operators are already doing about it. Kahi has put a QR code on every box, in Korea and overseas, since December 2023, and its verification screen presents number choices that change with every scan, requiring the customer to select the number printed above the code. That defeats an attacker who reprints the QR code alone, because the code by itself does not satisfy a challenge that moves.
It is a genuinely better design than a static code, and it deserves saying so. It also still resolves to ink on a box. Whoever can produce the box can produce the printed number, and in a trade where a separate supplier makes the boxes, the boxes are available.
The objections, briefly
Batch numbers were never meant to be anti-counterfeiting devices. Correct, and that is the argument, not a defence. The problem is that regulators, distributors and consumers now use them as though they were. HSA recalled by batch number. APR rebutted by batch number. It was the only handle either party had.
The brand can check its own records. APR did, and that is how it established it had no record of exporting those numbers to Singapore. Records tell you what you shipped. They cannot tell you what is in a jar on a shelf in another country, which is why the question stayed open.
A serial number would fix it. A serial number narrows the problem from a lot to a unit, which matters, but a printed serial is still printed. Thailand’s FDA invites patients to type the serial number from a box into its Track & Trace page. That confirms a number was issued. It does not confirm that the box in your hand is the box it was issued for.
Most product is genuine. Almost certainly true. The difficulty is that nobody downstream can tell which units are not, so a copied number borrows the credibility of every honest one.
What each party should be able to prove
For a brand, the test is the one APR was put to in public, through no fault of its diligence: if a regulator hands you a single unit, can you establish whether you made it? Not whether you shipped that lot. Whether you made that object.
For a regulator, the test is whether a recall notice can be acted on by the person holding the product. A notice that identifies goods by batch number is actionable only for as long as the batch number is unique, and the Medicube file is a record of what happens when it is not.
For a distributor, the test is whether you can separate your inventory from product you never imported, when both carry the same string. Two of the three Singapore sellers were cleared and all three now carry a per-batch laboratory testing requirement, created by goods the manufacturer says never passed through its own export channel.
For a pharmacist, installer or shopper, the test is simpler and currently unanswerable. You are holding one object. Everything printed on it was printed by whoever made it, including the number that is supposed to tell you who that was.
Aeroz, Making Authenticity Checkable. Visit aeroz.io to learn more & get in contact with our team via [email protected].
© 2026 Aeroz. All rights reserved.