Patent attorneys spend years training themselves to notice the difference between words that sound almost identical. A translator without that same instinct can undo that discipline in a single sentence, and the damage often isn’t visible until years later, in a courtroom, when it’s far too expensive to fix. 📄
The Illusion of Equivalence
Ordinary language tolerates a wide margin of error. If someone describes a wall as “adjacent” to a window when it’s technically “contiguous” with it, nobody’s day is ruined. Patent claims don’t work that way. Every noun, verb, and connector in a claim has been placed there deliberately, often after multiple rounds of negotiation with a patent examiner, to draw an exact boundary around what is protected and what is fair game for competitors.
When that claim crosses into another language, a translator facing two near-synonyms has to make a judgment call in a fraction of a second. Choose the more natural-sounding word, and the sentence reads beautifully. Choose it without understanding claim construction, and the boundary the attorney fought to establish can shift without anyone noticing until a dispute forces a side-by-side comparison of the original and the translated text.
A Rome Courtroom and Two Small Words
The clearest illustration of this risk didn’t come from a dramatic infringement battle. It came from a single connector. In the dispute between TEMA Technologies and Materials and Schluter System, heard before the Court of Rome, the entire disagreement centered on a European patent describing a foam core layer as moisture-resistant “and” moisture-proof. In the Italian translation filed for national validation, that connector became “or.” The shift from “and” to “or” was later found to have broadened rather than narrowed the scope of the claim, prompting the challenger to argue the patent should be invalidated outright. Biglanguage
The court didn’t ultimately strike the patent down, but the case is still cited by IP litigators today as a warning. Legal commentators have pointed to the July 2021 ruling as a clear example of how an incorrect Italian rendering of a European patent’s claim language can create real legal consequences for the patent holder, a risk that grows heavier in any jurisdiction that requires a full national-language validation after grant. One connector, chosen carelessly, put an entire granted right back on trial. Lexology
Where Precision Breaks: Five Term Pairs That Redefine a Claim
Patent drafters rely on a small vocabulary of scope-defining language that reads almost identically to a general audience but carries sharply different legal weight. The choice between “comprising” and “consisting of,” between “adjacent” and “contiguous,” between “approximately” and “substantially” affects what a claim covers and what it excludes, and a translator substituting a near-synonym for any of these terms can inadvertently narrow or widen the entire scope of protection. Global-lts
| Original Term | Risky Substitute | Legal Effect of the Swap |
|---|---|---|
| Comprising | Consisting of | Open claim becomes closed; excludes unlisted elements |
| Adjacent | Contiguous | Proximity requirement becomes a direct-contact requirement |
| Approximately | Substantially | Numerical tolerance shifts to a qualitative, less predictable standard |
| And | Or | Conjunctive requirement becomes an either/or condition |
| Shall | May | Mandatory step becomes optional, weakening enforceability |
None of these substitutions would raise an eyebrow in a general business document. In a patent claim, each one can move the line between what’s protected and what’s freely available to a competitor. 📊
Why Automated Tools Struggle With This Specific Trap
Statistical and AI-based translation engines are trained to reward fluency and naturalness. That’s precisely the problem here: a model optimizing for “how would a native speaker phrase this” will often reach for the more common word rather than the legally correct one, because the common word simply appears more often in its training data. The claim ends up reading smoothly in the target language while quietly drifting from the scope the original inventor secured. This isn’t a hypothetical edge case — it’s the exact failure pattern that scope-defining terms are most vulnerable to, since near-synonyms are, by definition, the substitutions a fluency-driven system is most likely to make without flagging them.
How Courts and Examiners Actually Read the Gap
Under the European Patent Convention, the authentic text of a European patent is the one filed in the language of proceedings before the EPO, but once translated into a contracting state’s official language for validation, if the scope of protection conferred by the translated patent is narrower than that conferred by it in the language of the proceedings, the translation shall be regarded as the authentic text in that state. In practice, this means that a careless translation doesn’t just risk confusion — it can legally become the operative document that a national court relies on, even when it fails to match what the applicant actually intended to protect. IP-PorTal
That is why the correction of a translation error is rarely simple after the fact. Federal Circuit rulings on claim correction consistently require that any fix be beyond reasonable debate, supported by the specification, drawings, and prosecution history, before a court will even consider adjusting the language. One November 2025 decision reinforced that judicial correction of a claim error remains available only where the proposed fix aligns with the claim structure, the figures, and the specification, and where no reasonable alternative correction exists — a high bar that a translation-driven error may not clear cleanly, especially when the mistake was never disclosed to the patent office in the first place. Morgan Lewis
🖼️
What a Careful Translation Process Actually Requires
A handful of practices consistently separate patent filings that hold up under scrutiny from those that don’t:
- Claim-by-claim terminology mapping before translation begins, so scope-defining words are locked to a single approved rendering across the entire filing 📝
- A dedicated glossary built from the specification itself, not a general bilingual dictionary, since general dictionaries are exactly where near-synonym substitutions originate
- A technical and legal review pass by someone who understands claim construction, not only fluency in both languages
- A side-by-side comparison of connectors, transition phrases, and modifiers (“and/or,” “comprising/consisting,” “approximately/substantially”) as a final checkpoint before filing
- Version control that preserves the reasoning behind each terminology choice, so future amendments or oppositions can be defended with a clear paper trail
The Word Nobody Double-Checks Is the One That Ends Up in Court ⚖️
Every patent portfolio is only as strong as its weakest translated clause, and that clause is rarely the one everyone worries about. Attorneys pore over inventive step arguments and prior art searches, but the small connecting words that hold a claim together often pass through translation with far less scrutiny than they deserve. The TEMA v. Schluter dispute shows how little it takes for that gap to become a genuine liability — not a mistranslated technical term, but a single three-letter word standing between “and” and “or.” Businesses that treat terminology consistency as a formality, rather than a legal safeguard, are the ones most likely to discover the difference the hard way. 🌐
source: https://new.epo.org/en/legal/epc/2020/a70.html
source: https://www.uspto.gov/web/offices/pac/mpep/s1412.html
source: https://www.lexology.com/library/detail.aspx?g=1c585888-8107-4897-aebf-3a1b2790c242