On this page
Why the translation can beat the originalHow grammatical number gets lostKorean has the same trap, built inA number check for any validation or national-phase translationQuick answersMost validation translations are filed in a hurry, checked by nobody who will ever litigate the patent, and forgotten. Then, years later, someone reads the Dutch claims in court and finds that “electrodes” became “an electrode”.
That single letter matters because of how Europe decides which text counts. The English, French or German text is the authentic one before the EPO. In a national court, it may not be, and the Netherlands is one of the places where the answer can turn against the patent owner.
Why the translation can beat the original
Article 70(3) of the European Patent Convention lets a member state treat its own-language translation as the authentic text when that translation confers narrower protection than the original. The exception is revocation: validity is still judged on the original text. In practice that asymmetry is the worst of both worlds for an owner. The patent can be attacked at full width and enforced only at the width of the translation.
The Dutch claims requirement is not optional. Under the London Agreement the Netherlands waived a full translation for patents granted in English, but it still requires the claims in Dutch, filed within three months of grant and with a small fee. Miss it and the patent has no effect there. So almost every European patent validated in the Netherlands carries a Dutch claim set that somebody translated quickly.
The risk is not theoretical. In 2024 the Kluwer Patent Blog reported a Dutch case in which the first-instance court limited a patent's scope based on the Dutch translation of the claims. The Court of Appeal took a different view, but the patentee had to go through an appeal to get there. That is the cost of a translation nobody reviewed.
The patent is attacked at the width of the original and enforced at the width of the translation.
How grammatical number gets lost
Plural-to-singular errors rarely look like errors. They look like tidy translation. Three patterns come up again and again:
“electrodes” → one electrode
Source: a plurality of electrodesTarget: an electrodeA claim meant to cover multi-electrode devices now reads on single-electrode ones only. A competitor with three electrodes is outside the literal wording.“the layers” → “the layer”
Source: wherein the layers comprise…Target: wherein the layer comprises…The antecedent was plural. The translation silently picks one of them, and the limitation now applies to a different thing.Open count made exact
Source: at least one sensorTarget: a sensorMany languages read a bare singular as “exactly one” more readily than English does. The open count disappears.None of these is a spelling mistake a proofreader would flag. They are choices, made by someone who did not know that number was the point of the claim. We see the same mechanism with conjunctions, covered in how a two-letter “and”/“or” error put a European patent at risk, and with statistics, in “average” vs. “median”.
Korean has the same trap, built in
For anyone filing in Korea, this Dutch problem should sound familiar. Korean nouns do not need a plural marker. A bare noun such as “sensor” can mean one sensor or several, and the optional plural ending is rarely used in technical writing. Number has to be expressed on purpose:
| English claim wording | What the Korean claim must express | What to check |
|---|---|---|
| a plurality of | Two or more, with the established claim term for a plurality | Not dropped in later references back to the same element |
| one or more | An explicit lower bound of one, open above | Not shortened to a bare noun “for flow” |
| at least one | An explicit minimum of one | Kept wherever the source repeats it |
| each of the | That the feature applies to every member individually | Distributive meaning kept, not read as a single collective |
A Korean translation that renders “a plurality of electrodes” correctly in claim 1 and then refers to a bare, unmarked “electrode” in claim 7 has created exactly the ambiguity the Dutch court had to resolve. KIPO examiners read the Korean text, and so will a Korean court. Korea does allow certain mistranslation corrections within the permitted periods, but it is a procedure with conditions, not a safety net. We look at those limits in legal nuances in Korean patent translation.
A number check for any validation or national-phase translation
- List every claim element introduced with a plural, “a plurality of”, “one or more”, “at least one” or “each”.
- Follow each one through every later reference (“the”, “said”, “each of the”) in all dependent claims.
- Confirm the translation keeps the count at every reference, not only at the first mention.
- Where the target language leaves number unmarked, add the explicit quantifier the source used, and record why.
- Compare the translated claims against the original once more after the final amendments. Claim sets copied from an earlier draft are where these errors hide.
Quick answers
Is the original-language text always the one that counts?
Before the EPO, yes. In a national court, a narrower national translation can count for infringement. Validity is still judged on the original.
Can a narrow validation translation be corrected?
Yes in the states that apply the rule, with conditions, and possibly with rights preserved for third parties who relied on the narrow version.
Does Korean have the same problem?
Yes. Number is optional in Korean nouns, so it has to be written in deliberately. See also ten common English-to-Korean patent translation mistakes.
Filing in Korea or validating in Europe?
We check grammatical number, conjunctions and antecedents claim by claim, and flag every place where the target language forces a choice. Ask for a claim review →