Drafting for the UPC: What Early Equivalents Decisions Tell Us About Claim Scope

Key takeaways

● The UPC's four-question equivalence test, first applied in Plant-e v Arkyne, makes technical-functional equivalence the dispositive threshold. If the substitute element does not perform the same function for the same result, equivalence fails before any other factor is considered.

● The majority of post-Plant-e decisions across Brussels, Mannheim, Paris, and The Hague have denied equivalence, suggesting the test is harder to satisfy than the first ruling might imply. However, another Hague decision (Washtower v Wasombouw) confirmed that equivalence can still succeed where the threshold is met.

● Deliberate use of restrictive claim language (e.g., "non-overlapping") has been held to defeat equivalence arguments. This is a direct warning to drafters about word choice.

● The description is always consulted when interpreting claims, not just when ambiguity exists, but it cannot narrow claims below their literal meaning.

● A Court of Appeal ruling on equivalence is still outstanding. Practitioners should draft for uncertainty.

The legal framework: built from parts, not from whole cloth

The UPC's approach to infringement sits at the intersection of three provisions that were never designed to work together as a unified doctrine.

Article 69(1) EPC directs that the extent of protection is determined by the claims, with the description and drawings used for interpretation. Article 1 of the Protocol balances fair protection for the patentee against reasonable legal certainty for third parties. Article 2 of the Protocol introduces equivalence, mandating that "due account" be taken of equivalent elements.

Article 24(1)(c) of the UPC Agreement incorporates all of this into UPC law. But the UPCA itself says nothing about how equivalence should be assessed — no test, no factors, no guidance. That gap is being filled by the local divisions.

The foundational claim construction decisions

Before equivalence enters the picture, the UPC had to establish how claims are interpreted in the first place. Three Court of Appeal decisions set the framework.

10x Genomics v NanoString (CoA, UPC_CoA_335/2023, 26 February 2024) established that the patent claim is the "decisive basis" for determining the scope of protection, not merely a starting point, but the anchor. It set out seven core interpretive principles under Article 69 EPC and the Protocol. Key among them: claims must be read from the skilled person’s perspective, using a technical rather than philological understanding. The same construction applies for both infringement and validity.

VusionGroup v Hanshow (CoA, UPC_CoA_1/2024, 13 May 2024) refined these principles with what has become the most-cited proposition in UPC claim interpretation: "Features are always interpreted in light of the claim as a whole." This means individual terms cannot be read in isolation — their technical function must be deduced from the claim context.

Meril v Edwards (CoA, UPC_CoA_464/2024, 25 November 2025) pushed the framework further, emphasizing that claim comparison must account for the "inventive concept" and "technical teaching" of the patent. Claims are compared as a whole, not dissected feature by feature.

Together, these decisions establish a claim construction regime that is purposive and technically grounded. That matters for equivalence: how broadly you read the literal claim determines how much room equivalence has to operate.

Plant-e v Arkyne: the first equivalence finding

In November 2024, the Local Division of The Hague issued the UPC's first decision finding infringement by equivalence. In doing so, it wrote the test that every subsequent court has applied or engaged with.

Plant-e Knowledge B.V. v Arkyne Technologies S.L. (LD The Hague, UPC_CFI_239/2023, 22 November 2024) concerned European Patent EP 2 137 782, covering microbial fuel cells powered by living plants. Plant-e's patent claimed a device where plant roots sit directly in an anode compartment alongside microorganisms. Bioo (the alleged infringer) deliberately engineered around this by separating the plants from the anode via a filter. This two-compartment structure avoided root damage to the anode while still allowing organic compounds to leach downward.

The court established a two-step infringement framework: assess literal infringement first; proceed to equivalence only if literal infringement is not established. For the equivalence step, it introduced a four-question test:

1. Technical equivalence. Does the variant solve essentially the same problem and perform essentially the same function to achieve essentially the same result?

2. Fair protection. Is extending protection proportionate to the patentee's contribution to the art, and would a skilled person find it obvious how to apply the equivalent element?

3. Legal certainty. Does the skilled person understand from the patent that the scope of the invention extends beyond the literal claim wording?

4. Novelty / Gillette-Formstein defence. Is the accused embodiment novel and inventive over the prior art?

The court answered all four questions affirmatively. It held that Plant-e had "established a new category of microbial fuel cells," warranting a "fairly broad scope of protection". On legal certainty, the court found that "the skilled person will understand that the variation is another way to obtain this result in a similar way".

The decision was seen as relatively patentee-friendly. Some commentators noted concern that it could grant exclusive rights covering "all feasible ways to achieve a stated result", potentially chilling design-around innovation.

After Plant-e: a pattern of denial

If Plant-e suggested a broad approach to equivalence, the majority of subsequent decisions have pushed back (though not universally).

LD Brussels in OrthoApnea S.L. (UPC_CFI_376/2023, 17 January 2025) denied equivalence, finding no functional equivalence on the facts. LD Mannheim in DISH Technologies v Aylo Premium (UPC_CFI_471/2023, 6 June 2025) denied equivalence and went further, suggesting that the UPC should develop its own harmonized test rather than borrowing from national doctrines. It ultimately left this question open, deciding the case on the basis that equivalence failed under all national tests.

LD Paris in N.J Diffusion v Gisela Mayer (UPC_CFI_363/2024, 1 August 2025) denied equivalence. A second LD The Hague decision in Advanced Cell Diagnostics v Molecular Instruments (UPC_CFI_187/2024, 18 November 2025) denied equivalence where the patentee had used deliberately restrictive claim language like "non-overlapping." The court held that the skilled person would understand such narrow terms as reflecting a deliberate technical choice.

The one exception is Washtower v Wasombouw/Bega (LD The Hague, UPC_CFI_479/2025, 17 September 2025), where the court applied the Plant-e framework and found infringement by equivalence. The test is not insurmountable, but the rarity of success underscores how demanding it is.

The emerging pattern is clear: functional equivalence operates as a gate. If the substitute element does not perform essentially the same function to achieve essentially the same result, the analysis stops. Courts are not reaching the fair protection or legal certainty questions in most cases.

What this means for claim drafting

These decisions are not just of academic interest. They create concrete drafting pressures that practitioners should be responding to now.

Functional language is your friend — up to a point

Since technical-functional equivalence is the dispositive threshold, claims that describe elements by function rather than structure are better positioned to capture equivalents. Consider: "a separation element configured to permit passage of organic compounds" is more resilient than "a filter". Functional language pre-frames the equivalence inquiry in the patentee’s favour.

But there are limits. Overly broad functional language risks invalidity objections. Under the NanoString principles, functional terms must be interpreted in light of the description and drawings. A functional claim term not supported by concrete embodiments may be read narrowly or challenged for insufficiency.

Word choice has consequences you cannot undo

The November 2025 Hague decision on "non-overlapping" is a stark warning. When a drafter deliberately chooses a restrictive term, courts will hold the patentee to it. That applies not just at the literal infringement stage, but at the equivalence stage as well. The skilled person is treated as understanding that the patentee's precise language reflects a deliberate technical choice.

Drafters must consider claim language not only for what it literally covers, but for what it signals about intended scope. Restrictive modifiers, specific numerical ranges, and structural limitations all carry equivalence risk. Where possible, use the broadest defensible formulation in the independent claim and save narrower alternatives for dependents.

Your description will be read — make sure it helps

Under VusionGroup and the NanoString principles, the description and drawings are always consulted when interpreting claims. Not just when the claim language is ambiguous, but always. At the same time, FUJIFILM v Kodak (LD Düsseldorf, UPC_CFI_355/2023, 28 January 2025) established that the description cannot narrow claims below their literal wording. It held that:

"Art. 69 EPC and its Protocol do not provide a justification for excluding what is literally covered by the terms of the claims by a narrowing claim construction based on the description or drawings."

This creates a one-way ratchet: the description can inform and contextualise, but it cannot cut back what the claims plainly cover. The practical implication is clear. The description should support the broadest reasonable interpretation of claim terms while remaining consistent with them. Describe multiple embodiments. Explain the technical problem broadly enough to encompass foreseeable variations. Avoid language that inadvertently limits claim terms to a single implementation.

Prior art constrains equivalence

The fourth question (the Formstein/Gillette defence) prevents patentees from using equivalence to capture scope they could not have claimed over the prior art. When writing claims, consider whether your claimed scope, including the equivalence penumbra, is novel and inventive. If not, the equivalence argument will fail.

The Washtower v Wasombouw decision (UPC_CFI_479/2025) confirmed this, dismissing the defendants' prior art arguments under this fourth question before upholding equivalence.

Prosecution history is a growing risk

The UPC’s position on file wrapper estoppel remains uncertain, but early indications suggest prosecution history may constrain equivalence arguments. If a claim was narrowed during prosecution to overcome prior art, recapturing that surrendered scope through equivalence will be difficult. Practitioners should maintain clean prosecution histories and avoid concessions that could later defeat equivalence.

How AI-assisted drafting changes the calculus

The drafting pressures above share a common feature: they require practitioners to balance competing considerations simultaneously. Functional breadth versus invalidity risk. Word choice signalling versus technical accuracy. Description support versus unintended limitation. Equivalence penumbra versus prior art. These are the kind of multi-variable problems where AI-assisted drafting tools can reduce error rates and improve coverage.

Solve Intelligence offers a patent copilot specifically built for this workflow, with capabilities that map directly onto the challenges the UPC’s early equivalence case law creates.

Generating alternative embodiments to support claim breadth

The Plant-e four-question test makes functional equivalence dispositive, and the description is always consulted to inform that analysis. Practitioners need descriptions that disclose multiple embodiments at varying levels of abstraction to support the broadest reasonable claim interpretation. Solve Intelligence’s drafting copilot generates alternative embodiments and applications from an initial invention disclosure, helping populate the description with the variation that supports equivalence arguments downstream. A description disclosing only one implementation of a claimed function invites courts to read it narrowly. AI-assisted generation of alternatives directly mitigates this risk.

Catching restrictive language before it calcifies

The November 2025 Hague decision on "non-overlapping" illustrates how a single word choice can foreclose equivalence entirely. The difficulty is that restrictive modifiers are often introduced unconsciously. A drafter reaches for precision and inadvertently signals limitation.

Solve Intelligence’s patent review tool (Solve Review) analyses draft applications against EPO Guidelines and other frameworks, flagging language that may unintentionally narrow scope. No tool can predict how a UPC division will interpret a specific term years from now. But systematic review against known standards catches the most obvious traps before they enter the file.

Preserving scope during prosecution

Prosecution history estoppel carries particular force at the UPC, where amendments may constrain equivalence arguments. Solve Intelligence’s office action response tool addresses the core tension: overcoming examiner objections while maintaining the broadest possible scope.

It parses claim limitations against cited prior art, identifies unmapped elements and overbroad examiner interpretations, and proposes targeted amendments that address rejections without conceding scope. For practitioners managing large portfolios that may face UPC enforcement, maintaining clean prosecution histories at scale is a meaningful strategic advantage

Mapping equivalence risk with claim charts

Solve Intelligence’s claim charting tool, launched in late 2025, uses semantic matching and synonym detection to identify when two disclosures describe functionally equivalent elements in different language. This is directly relevant to pre-filing equivalence analysis. Practitioners can map draft claims against known competitor products and prior art to assess whether their equivalence penumbra captures anticipated design-arounds or collides with prior art that would trigger the Formstein defence. Running this analysis before filing allows claims to be adjusted while adjustment is still possible.

What AI does not replace

None of this substitutes for the practitioner's judgment on claim strategy. Key doctrinal questions remain matters of legal and technical expertise. How much functional language will a particular examining division tolerate? Will a specific word choice be treated as a deliberate limitation? How might a Court of Appeal recalibrate the Plant-e test?

What AI-assisted drafting does is compress the mechanical dimensions of the problem: generating embodiments, checking language, mapping prior art, preserving scope during prosecution. The time recovered is time that can be redirected to the strategic questions that the UPC's still-developing equivalence doctrine makes genuinely difficult.

The road ahead

No Court of Appeal decision on equivalence exists yet. When it comes (and it will) it may confirm the Plant-e four-question test, modify it, or replace it with a UPC-specific framework as Mannheim has suggested. The current local division consensus around functional equivalence as a threshold gate may survive or be recalibrated.

What will not change is the underlying tension built into the Protocol on Article 69 EPC: fair protection for the patentee versus legal certainty for third parties. Every equivalence decision negotiates between these two poles. The UPC’s early case law suggests it is resolving that tension cautiously: willing to recognise equivalence in principle, but demanding rigorous proof of functional equivalence in practice.

For patent drafters, the message is straightforward. Do not rely on equivalence to rescue narrow claims. Draft the scope you need into the claims themselves; using functional language, multiple levels of abstraction, and descriptions that support breadth rather than constrain it. The doctrine of equivalents at the UPC is a safety net, not a strategy. And the early decisions make clear it is a net with a tight weave.

Frequently Asked Questions

What is the UPC's test for infringement by equivalence?

The Plant-e v Arkyne decision (LD The Hague, November 2024) set a four-question test: technical-functional equivalence, fair protection, legal certainty, and a novelty/Formstein defence. Every subsequent local division has applied or engaged with it, but no Court of Appeal ruling has confirmed it yet. Most cases fail at the first question. If the substitute does not perform the same function for the same result, the analysis stops there.

Can restrictive claim language prevent equivalence arguments later?

Yes. The November 2025 Hague decision held that choosing a narrow term like "non-overlapping" signals a deliberate technical choice. That forecloses equivalence at enforcement. Use the broadest defensible formulation in independent claims. Save narrower alternatives for dependents. Word choice at drafting carries consequences that equivalence cannot undo.

How can Solve Intelligence help practitioners draft claims resilient to the UPC's equivalence framework?

Solve's Patent Copilot generates alternative embodiments to support claim breadth, flags restrictive language that may narrow scope unintentionally, and maps draft claims against prior art to test whether the equivalence penumbra holds up or triggers a Formstein defence. It handles the mechanical complexity so practitioners can focus on the strategic questions that matter.

‍