Evidence-grounded memo prepared for the UPC's feedback-on-improvements consultation. Phase 1: identification of weakness clusters from the body of decided case law. Proposed rule amendments are deliberately held for Phase 2.
After three years of operation the Unified Patent Court has produced a substantial body of procedural case law. That case law is the most reliable diagnostic instrument available for the Rules of Procedure (RoP): where a rule is well-drafted, it is applied and disappears from view; where it is silent, vague, or internally inconsistent, the divisions are forced to invent the missing test — and they frequently invent different tests. The recurring German term the courts themselves use for this is "planwidrige Regelungslücke" — an unplanned regulatory gap — and they fill it by analogy, case by case, division by division.
This memo mines ~1,950 classified UPC decisions to surface the rules where that gap-filling is most concentrated and most divergent. The single clearest signal is frequency of contested interpretation: R.262/R.262A (confidentiality and public access) is litigated more than twice as often as any other procedural rule — 110 distinct decisions across 26 different judicial formations — yet the rule text supplies almost none of the tests the courts apply.
The eleven weakness clusters, each with a one-line problem statement:
The meta-pattern across all eleven: the RoP was drafted as a framework that assumes judicial completion, but it does not tell the divisions how to complete it, and there is no fast appellate mechanism to harmonise first-instance gap-filling before it ossifies into divergent local practice. The result is forum-dependent procedure on questions — access, security, fees, stays, extensions — that ought to be uniform and foreseeable across a single Court.
Data source. data.ai_judgment_notes in the yoUPC.org database — AI-classified annotations (holdings, legal principles, provision interpretations, key quotes, case outcomes) over ~1,950 UPC decisions, cross-referenced to data.judgments metadata (court, date, parties) and data.judgment_markdown_content (full text). Rule text quoted verbatim from the youpc-laws corpus (UPCRoP.NNN.md, EN).
Signals mined. 1. Frequency of contested interpretation — provision_interpretations ranked by how often each rule is interpreted, and by how many distinct judicial formations interpret it. High counts indicate the wording or its application is repeatedly contested. 2. Cross-division divergence — the same rule reaching different conclusions in different divisions. 3. Judicial gap-filling — decisions that cite a rule but apply a test the rule text does not contain (often expressly labelled a "Regelungslücke" / "unplanned gap"). 4. Explicit rule criticism — holdings and key quotes containing phrases such as "the Rules do not provide", "is silent", "no guidance", "open question", "lacuna / Regelungslücke", "does not address/specify".
Each candidate cluster was then deep-dived: all interpretations for the rule pulled and grouped by court, supporting holdings and verbatim quotes retrieved, and the actual rule wording read against the case law. Every case number cited below was verified to exist in the database.
Most-litigated procedural rules (distinct decisions interpreting each; the count is of contested interpretation, not mere citation):
| Rule | Subject | Decisions | Formations |
|---|---|---|---|
| R.262 / R.262A | Confidentiality & public access | 110 | 26 |
| R.9 | Case management & time periods | 72 | 14 |
| R.263 | Amendment of case | 41 | 15 |
| R.158 | Security for costs | 41 | 15 |
| R.265 | Withdrawal | 32 | 12 |
| R.211 | Preliminary injunctions | 32 | 11 |
| R.30 | Amendment of patent / auxiliary requests | 25 | 14 |
| R.370 | Court fees & refunds | 24 | 12 |
| R.295 | Stay of proceedings | 20 | 10 |
| R.190 | Order to produce evidence | 19 | 7 |
| R.333 | Review of case-management orders | 18 | 9 |
| R.19 | Preliminary objection | 18 | 8 |
| R.220 | Appeal / discretionary review | 15 | 5 |
| R.151 | Cost decision | — | — |
Scope and its limits. This analysis is judgment-derived. It surfaces rules that have generated contested case law. A rule can be badly drafted yet rarely litigated — such rules will not surface here even though they may merit reform (for example, rules governing procedural steps that parties simply comply with, or rules whose defects deter litigation altogether). The frequency table is therefore a map of where disputes have arisen, not an exhaustive audit of drafting quality. Where the evidence base for a claim is thin, this is flagged in the cluster. The analysis is confined to the RoP itself; comparison with EPC/national procedure and academic commentary is out of scope (Phase 2).
The rule. R.262.1(b): "written pleadings and evidence, lodged at the Court and recorded by the Registry shall be available to the public upon reasoned request to the Registry; the decision is taken by the judge-rapporteur after consulting the parties." R.262.6 frames the access standard only as: "The Court shall allow the Application unless legitimate reasons given by the party concerned for the confidentiality of the information outweigh the interest of the applicant to access such information." For confidentiality clubs, R.262A.6 fixes only a floor: the protected circle "shall be no greater than necessary … and shall include, at least, one natural person from each party and the respective lawyers or other representatives."
What courts are doing. The text says access follows a "reasoned request" but never defines it. The Munich Central Division built the gap-filler in the foundational Ocado/AutoStore line: a reasoned request must state which documents are sought, the purpose, and why access is necessary, and the applicant must give a "concrete, verifiable, legitimate reason" — more than "any (fictitious) reason." The Court of Appeal canonised that definition in UPC_CoA_404/2023 (Ocado v AutoStore) and added an unwritten requirement that public-access applicants be represented. On scope, the divisions read R.262.1(b) to cover only substantive party pleadings and evidence — excluding registry correspondence, formality checks and EPO communications — a limitation found nowhere in the text. On confidentiality clubs, R.262A.6 fixes only a floor; the Court of Appeal invented a three-factor test (role, relevance, trustworthiness) to size the club and decide individual access (UPC_CoA_621/2024 — Daedalus Prime v Xiaomi), and held that employee status alone cannot exclude a nominee. Hamburg (UPC_CFI_169/2024) declared an unwritten UPC-wide minimum: a court may not order an attorneys'-eyes-only club.
Divergence. There is a real split on the R.262.1(b) threshold. The Munich CD (UPC_CFI_1/2023, UPC_CFI_75/2023 — Astellas) and the CoA (Ocado) demand a "concrete and verifiable, legitimate reason" and reject generic purposes such as education or training — a substantive gate. The Nordic-Baltic Regional Division (UPC_CFI_8/2023) and Hamburg (UPC_CFI_151/2024) instead apply a presumption-of-access "credible explanation" standard — the request "shall be approved unless it is necessary to keep the information confidential" — treating the requirement as essentially formal. On club composition, Hamburg (UPC_CFI_22/2023) once treated the one-natural-person floor as a default displaceable by party agreement, while Hamburg (UPC_CFI_169/2024) and Düsseldorf (UPC_CFI_355/2023) treat it as a mandatory "ground rule of paramount importance."
Evidence (cited decisions): - UPC_CFI_1/2023 (Munich CD) — R.262.1(b) requires a concrete, verifiable, legitimate reason; scope limited to party pleadings/evidence, excluding registry/EPO correspondence. - UPC_CFI_14/2023 (Munich CD) — "written pleadings or evidence" means documents relating to the substance of a party's case, not registry correspondence or formality checks. - UPC_CFI_75/2023 (Munich CD, Astellas) — reasoned request must state documents, purpose and necessity; education/training is not a legitimate reason. - UPC_CoA_404/2023 (CoA, Ocado v AutoStore) — canonical definition of "reasoned request" plus the invented representation requirement. - UPC_CFI_8/2023 (Nordic-Baltic RD) — presumption-of-access reading; a credible explanation suffices — a markedly lower bar than Munich CD. - UPC_CFI_169/2024 (Hamburg LD) — invents a UPC-wide minimum: at least one natural person per party must access confidential material; no attorneys'-eyes-only clubs. - UPC_CoA_621/2024 (CoA, Daedalus Prime v Xiaomi) — three-factor club-access test (role, relevance, trustworthiness). - UPC_CFI_355/2023 (Düsseldorf LD) — the one-natural-person floor as a "ground rule of paramount importance"; per-redaction substantiation required.
Practical impact. A member of the public or competitor cannot predict from the rule text whether a generic purpose ("monitor UPC case law") will clear the access bar — it depends on whether the assigned division follows the strict Munich/CoA line or the Nordic-Baltic presumption-of-access line. Patentees seeking confidentiality face the inverse uncertainty: the bare R.262A text suggests broad discretion to limit the club, but the courts have read in a hard floor (at least one natural person, no AEO) and a multi-factor proportionality test, so attorneys'-eyes-only protective orders familiar from other jurisdictions will fail. Drafting a confidentiality application now requires anticipating unwritten, division-specific tests rather than the rule itself.
Proposed direction (signal only). Codify the "documents + purpose + necessity" content of a reasoned request and the applicable access standard in R.262, and codify the three-factor club-sizing test and the one-natural-person / no-AEO floor — all of which the case law has already settled — in R.262A.6.
Confidence: strong. 8+ distinct decisions across the Munich CD, Nordic-Baltic RD, Hamburg, Düsseldorf, Mannheim and Milan LDs and the CoA; the threshold divergence is directly evidenced, not inferred.
The rule. R.158.1: "At any time during proceedings, following a reasoned request by one party, the Court may order the other party to provide, within a specified time period, adequate security for the legal costs and other expenses … Where the Court decides to order such security, it shall decide whether it is appropriate to order the security by deposit or bank guarantee." R.158.2 adds only a right to be heard and a cross-reference to enforcement. That is the entire substantive content — "adequate security," "reasoned request," and "may."
What courts are doing. The void was filled with a uniform two-step burden-shifting test imported from the Central Division (UPC_CFI_252/2023 — NanoString v Harvard) and elevated by the Court of Appeal on 17 September 2024 in the Audi v Network System Technologies line (UPC_CoA_217/2024 and parallels): the applicant must show, on specific facts, a "legitimate and real concern" that a costs order would be unrecoverable and/or only enforceable in an "unduly burdensome way." On amount, the rule is silent: Lisbon (UPC_CFI_850/2026 — Transsion v Ericsson), The Hague and Paris CD anchor at 50% of the recoverable-costs ceiling, while Düsseldorf (UPC_CFI_100/2024 — Ona Patents v Google) rejects the ceiling as an automatic anchor and applies R.152 proportionality instead. On form, the CoA (UPC_CoA_301/2024 — ICPillar v ARM) and Munich (UPC_CFI_617/2025) read "deposit or bank guarantee" restrictively — US-bank guarantees and ATE insurance are inadequate. For release/revocation the rule says nothing: Munich CD and Düsseldorf graft R.352.2 on by analogy, and Paris CD invents an "implied power" to revoke and vary (UPC_CFI_164/2024 — Microsoft v Suinno).
Divergence. Multi-axis and real: 1. Relative financial position — an early Munich LD line treated comparative wealth as a factor; the CoA (Audi line, reaffirmed in UPC-COA-0000001/2026 — Adobe v KEEEX) holds it is not a criterion. 2. Amount — Lisbon / The Hague / Paris CD anchor at 50% of the ceiling; Düsseldorf rejects the ceiling as an automatic anchor. 3. PI proceedings — Düsseldorf says R.158 has no scope in urgent proceedings; Hamburg applies it in PI. 4. Form — CoA / Munich reject US-bank guarantees and ATE insurance. 5. Release / revocation — each court borrows a different hook (R.352.2 analogy vs. Paris CD's "implied power" vs. the CoA's R.335/R.336 case-management route). 6. Claimant-side — the CoA (UPC_CoA_393/2025 — AorticLab v Emboline) holds Art.69(4) UPCA gives no basis to secure costs at the claimant's request; a defendant is never obliged to provide security (confirmed UPC_CFI_525/2024 — Visibly v Easee).
Evidence (cited decisions): - UPC_CFI_252/2023 (Munich CD, NanoString v Harvard) — established the two-factor recoverability/enforceability test; held R.352.2 is the basis for releasing a R.158 security because the rule has none. - UPC-COA-0000001/2026 (CoA, Adobe v KEEEX) — relative financial position is not a R.158 criterion. - UPC_CoA_301/2024 (CoA, ICPillar v ARM) — "adequate security" assessed substantively; US-bank guarantees and insurance may fall short. - UPC_CFI_164/2024 (Paris CD, Microsoft v Suinno) — invented an "implied power" to revoke/vary on changed circumstances; amount at 50% of the ceiling. - UPC_CFI_850/2026 (Lisbon LD, Transsion v Ericsson) — amount anchored at 50% of the recoverable-costs ceiling. - UPC_CFI_100/2024 (Düsseldorf LD, Ona Patents v Google) — the ceiling is a safeguard against over-recovery, not an automatic starting point. - UPC_CoA_393/2025 (CoA, AorticLab v Emboline) — no legal basis to order security at the claimant's request. - UPC_CFI_617/2025 (Munich LD) — form limited to bank guarantee or deposit; ATE insurance is not an alternative form.
Practical impact. Whether a foreign (especially non-EU) or thinly-funded claimant must post security, how much, in what form, and whether it can ever be released or reduced, depends heavily on the division seised — a forum-shopping vector at the case-management stage. Defendants posting US-bank guarantees or ATE insurance risk rejection and a R.355 default. And because the CoA confirms only claimants can be ordered to provide security, defendants are left without the reciprocal protection the rule's "one party"/"other party" wording superficially seemed to offer.
Proposed direction (signal only). R.158 should codify the CoA's recoverability/enforceability criteria, an express quantum anchor, the accepted forms of adequate security, and explicit release/variation powers — rather than leaving each to analogy and "implied power."
Confidence: strong. 40+ interpretations across the CoA, four Central Division sections and at least nine Local Divisions; every gap axis is supported by a distinct cited decision, with the CoA's 17 September 2024 Audi line as the harmonising anchor.
The rule. R.190.1: "Where a party has presented reasonably available and plausible evidence in support of its claims and has, in substantiating those claims, specified evidence which lies in the control of the other party or a third party, the Court may on a reasoned request by the party specifying such evidence, order that other party or third party to produce such evidence. For the protection of confidential information the Court may order that the evidence be disclosed to certain named persons only …"
What courts are doing. R.190.1 sets out qualitative ingredients — "reasonably available and plausible evidence," evidence "specified" and "in the control of" the other party, a "reasoned request" — but nowhere defines a relevance or proportionality threshold: how relevant the sought evidence must be, how specifically it must be identified, or when a request tips from legitimate to abusive. Munich (UPC_CFI_208/2023) held expressly that "the RoP contains no rule" ("Die VerfO enthält keine Regel…") requiring conclusive proof of relevance before production is ordered, and separately that the RoP gives the court no power to compel a producing party to file specific confidentiality motions — so the interplay with R.262A is itself unspecified. The Hague (UPC_CFI_327/2024) and Brussels (UPC_CFI_1357/2025 — GC Aesthetics v Establishment Labs) crystallised a "four cumulative conditions" framework (threshold evidence; specification + control; confidentiality protection; two-stage proportionality) that appears nowhere in the rule's words. To police the missing threshold, divisions invented the "fishing expedition" / "Ausforschung" doctrine (UPC_CFI_216/2023 Mannheim; UPC_CFI_450/2024 Munich), rejecting requests for "all" agreements or requests identical across parallel proceedings. The CoA (UPC_CoA_298/2024) layered on a "strict necessity, relevance and proportionality" standard borrowed by analogy from EU Directives 2004/48 and 2014/104 (PACCAR, RegioJet) — confirming the rule itself supplies no native test.
Divergence. No clean doctrinal split — the divisions converge on a judicially-constructed framework — but the source of the threshold differs (CoA imports EU-directive proportionality; CFI divisions build a four-condition test). The live disagreement is over staging in SEP/FRAND cases: Mannheim (UPC_CFI_210/2023) defers FRAND-comparable production until after reply, whereas The Hague (UPC_CFI_327/2024) runs the R.190 branch on an accelerated track ahead of the Statement of Defence.
Evidence (cited decisions): - UPC_CFI_208/2023 (Munich LD) — express gap-finding: no rule requires conclusive proof of relevance; no power to compel confidentiality motions. - UPC_CFI_327/2024 (The Hague LD) — invented the "four cumulative conditions" test; confidentiality routed through R.262A/R.262.2. - UPC_CoA_298/2024 (CoA) — imported "strict necessity, relevance and proportionality" from CJEU directive case law by analogy. - UPC_CFI_216/2023 (Mannheim LD) — requests for "all" SEP licences are inadmissible fishing expeditions. - UPC_CFI_450/2024 (Munich LD) — volume and identical-across-proceedings requests indicate a fishing expedition; burden on the applicant. - UPC_CFI_550/2024 (Munich LD) — no preclusive cut-off barring late-designated evidence; production power applies at any stage. - UPC_CFI_1357/2025 (Brussels LD, GC Aesthetics v Establishment Labs) — restated R.190 as four cumulative conditions including two-stage proportionality. - UPC_CFI_65/2024 (Munich LD) — R.190 cannot be used to find a new infringement theory after the first proves untenable — a judicially-read purpose limit.
Practical impact. Litigants cannot predict whether a production request will succeed, because the operative threshold lives entirely in divergent case law. SEP/FRAND applicants face especially unpredictable timing. The R.190/R.262A confidentiality interplay is undefined, so a producing party has no rule-based guidance on whether and when it must file confidentiality motions — Munich confirmed the court cannot even compel it to.
Proposed direction (signal only). R.190.1 should codify the judicially-developed threshold — a relevance/plausibility standard short of conclusive proof, an express anti-fishing specificity requirement, a proportionality criterion, and a cross-reference to R.262A for confidentiality at production.
Confidence: strong. 15+ cited cases across Munich, Mannheim, The Hague, Brussels and the CoA, with two verbatim German gap-statements from UPC_CFI_208/2023 directly evidencing the wording gap.
The rule. R.211.2: "In taking its decision the Court may require the applicant to provide reasonable evidence to satisfy the Court with a sufficient degree of certainty that the applicant is entitled to commence proceedings …, that the patent in question is valid and that his right is being infringed, or that such infringement is imminent." R.211.4: "The Court shall have regard to any unreasonable delay in seeking provisional measures." R.30 — which supplies auxiliary requests — applies on its terms only to main proceedings.
What courts are doing. R.211 is skeletal and the courts have judicially completed it on at least four axes. (1) Degree of conviction: R.211.2 says only "sufficient degree of certainty" with no quantum; the CoA (UPC_CoA_1/2024, following NanoString) and the divisions read in a "more likely than not / balance of probabilities" standard for validity and infringement (UPC_CoA_297/2024 lifting a PI on that basis). (2) Auxiliary requests: R.30 supplies them for main proceedings but the PI framework is silent. Lisbon (UPC_CFI_317/2024) held R.30 cannot apply by analogy in PI — auxiliary requests are "incompatible with the summary, non-merits, provisional character" of PI; Milan (UPC_CFI_380/2024) and Munich (UPC_CFI_693/2025) agree. The Court of Appeal (UPC_CoA_382/2024) expressly flagged that "the allowability of auxiliary requests in PI proceedings has not yet been decided by the Court of Appeal," deferring the question. (3) Urgency clock: R.211.4 names "unreasonable delay" but no start date; the courts invented one — delay runs from when the applicant knew or should have known enough to file a promising application (UPC_CFI_1927/2025 Düsseldorf). (4) Mootness: R.211 is silent, so courts apply R.360 by analogy (UPC_CFI_130/2024 The Hague).
Divergence. Genuine open splits: (a) auxiliary requests in PI — Lisbon/Milan/Munich flatly reject them; the CoA has not decided (UPC_CoA_382/2024 deferred), and a later CoA order admitted one under R.222 — first instance and the developing CoA line are not aligned; (b) ex-parte prior announcement before penalty enforcement — Düsseldorf (UPC_CFI_213/2025) grants precautionarily while expressly "noting Mannheim's contrary view" — an acknowledged split.
Evidence (cited decisions): - UPC_CFI_317/2024 (Lisbon LD) — R.30 cannot apply by analogy in PI; also held the applicant must state the date it became aware of infringement (the urgency clock is unwritten). - UPC_CoA_382/2024 (CoA) — allowability of auxiliary requests in PI has not yet been decided; question deferred — confirms the gap is open. - UPC_CFI_380/2024 (Milan CD) — auxiliary request to amend not admissible in PI given expediency/imminence. - UPC_CFI_693/2025 (Munich LD) — R.211.2 refers to "the patent concerned" (granted form); R.30 claim amendment not applicable in PI. - UPC_CoA_1/2024 (CoA) — defines "sufficient degree of certainty" as "more likely than not on the balance of probabilities" — a judge-made standard. - UPC_CFI_130/2024 (The Hague LD) — R.211 silent on mootness; courts fill the gap (R.360 applied by analogy elsewhere). - UPC_CFI_213/2025 (Düsseldorf LD) — prior-announcement-before-penalty question is "open"; grants precautionarily, noting Mannheim's contrary view. - UPC_CFI_1927/2025 (Düsseldorf LD) — urgency assessed from when the applicant knew/should have known — an entirely judge-built test.
Practical impact. A patentee preparing a PI cannot tell from R.211 what standard of proof applies, whether it may file fallback claim sets if validity looks shaky, or when its urgency clock started ticking. Because most divisions refuse auxiliary requests in PI, a patentee who fears its granted claims are vulnerable faces an all-or-nothing bet, and amending may itself signal that the granted patent is "probably not valid" (Munich, UPC_CFI_693/2025). The unwritten urgency clock and the open ex-parte question create enforcement traps where a procedurally correct order can still be attacked.
Proposed direction (signal only). Amend R.211 to codify the "more likely than not" conviction standard, define the urgency-clock start (knowledge/diligence date) already used by the CoA, and state whether and on what conditions a limited number of auxiliary requests may be filed in PI.
Confidence: strong. Multiple corroborating cases on each sub-issue from the CoA and several divisions; the auxiliary-request gap is confirmed open by the CoA itself (UPC_CoA_382/2024).
The rule. R.9.2: "The Court may disregard any step, fact, evidence or argument which a party has not taken or submitted in accordance with a time limit set by the Court or these Rules." R.9.3: "… on a reasoned request by a party, the Court may: (a) extend, even retrospectively, a time period referred to in these Rules or imposed by the Court; and (b) shorten any such time period."
What courts are doing. The rule says the Court "may" extend "even retrospectively" on a "reasoned request" but supplies no criterion for when discretion should be exercised, no deadline by which a retrospective request must be lodged, and no standard for R.9.2 disregard of late material. A near-uniform CFI line (Paris CD UPC_CFI_412/2023, UPC_CFI_189/2024; Düsseldorf UPC_CFI_355/2023; The Hague UPC_CFI_1791/2025) reads in: extensions are an "exceptional remedy" in a "strict deadline regime," available only on evidence that timely compliance was "objectively impossible or very difficult" through no fault of the party, and limited to the minimum extra time — and a requirement to apply "as soon as the difficulty is apparent." None of this appears in the text. On the silence itself, the CoA (UPC-COA-0000037/2026; UPC-CoA-61/2026) reads the absence of a stated limit as positively permitting a request filed after expiry, and even endorses re-reading another application as containing an implicit R.9.3(a) request.
Divergence. (1) On the retrospective-request silence: the CoA treats the absence of any time limit as permitting late-filed requests, discretion "solely for the court," while CFI divisions graft on a self-policing limit ("as soon as the difficulty is apparent" — Paris). (2) On strictness: most divisions demand "objective impossibility / exceptional circumstances" (Paris, The Hague, Milan), but Munich (UPC_CFI_127/2024) holds consensual extensions are "as a rule" granted, and Düsseldorf (UPC_CFI_1928/2025) admits late PI submissions on bare R.9.1 discretion — a softer line.
Evidence (cited decisions): - UPC-COA-0000037/2026 (CoA, standing judge) — R.9.3(a) silence permits a retrospective request filed after the deadline; no time limit; discretion "solely for the court." - UPC-CoA-61/2026 (CoA) — R.9.3(a) retroactive extension covers both an order issued post-expiry and a request lodged post-expiry. - UPC_CFI_412/2023 (Paris CD) — invents the missing test: objective impossibility/extreme difficulty not attributable to the party; application "as soon as the difficulty is apparent." - UPC_CFI_189/2024 (Paris CD) — R.9.3(a) used "with caution and in justified exceptional cases." - UPC_CFI_355/2023 (Düsseldorf LD) — a two-prong gloss: attempted timely filing plus prompt upload after expiry. - UPC_CFI_1791/2025 (The Hague LD) — R.9.3 a "narrow discretionary power" exercised "with caution" — none of this stated in the rule. - UPC_CFI_127/2024 (Munich LD) — divergent softer line: consensual extensions granted "as a rule." - UPC_CFI_869/2025 (Düsseldorf LD) — R.9.2 disregard of late material is "substantive" and cannot be done by the judge-rapporteur as case management — a gap on who exercises the power.
Practical impact. Litigants face an unpredictable, division-dependent landscape on both extensions and late submissions. Paris/The Hague demand evidenced objective impossibility; Munich treats consensual extensions as routine. The CoA's reading that a retrospective request has no deadline is a trap for the unwary opponent — relief can be sought, and granted, long after expiry, undermining reliance on a missed deadline.
Proposed direction (signal only). R.9.3(a) should state an outer limit for lodging a retrospective request and codify the exceptional-circumstances criterion the courts have already invented; R.9.2 should specify the deciding body.
Confidence: strong. R.9 is the second most-cited rule (72 cases); ~50 distinct interpretations across the CoA, four Central Division seats and at least seven Local Divisions, with multiple cases on each contested point.
The rule. R.263.2: leave "shall not be granted if … the party seeking the amendment cannot satisfy the Court that: (a) the amendment in question could not have been made with reasonable diligence at an earlier stage; and (b) the amendment will not unreasonably hinder the other party …" R.263.3: "Leave to limit a claim in an action unconditionally shall always be granted." By contrast R.265.2 requires the Court on withdrawal to "(c) issue a cost decision in accordance with Part 1, Chapter 5." R.263 contains no equivalent cost-regulation clause.
What courts are doing. The rules overlap on the commonest move in litigation — dropping part of a case — but draw the line and its cost consequence differently, and neither rule says where a partial limitation ends and a partial withdrawal begins. Milan (UPC_CFI_472/2024) reads R.263.3 expansively to cover dropping one of several patents, and expressly notes that "differently from rule 265 RoP, rule 263 RoP does not provide the regulation of the proceeding costs, because the proceedings continue" — leaving the cost of the abandoned part to the final R.118 decision. Düsseldorf (UPC_CFI_457/2023) defers the characterisation entirely. On the leave test itself, divisions split on strictness: Munich (UPC_CFI_41/2024) treats limbs (a) and (b) as independent, cumulative and discretion-free; Munich CD (UPC_CFI_252/2023) invents bright-line cut-off dates that reverse the burden; the CoA (UPC_CFI_44/2024) and The Hague (UPC_CFI_424/2023) read "reasonable diligence" generously and require courts to weigh mitigations before refusing.
Divergence. (1) The R.265 cost decision — mandatory ex officio vs. request-based: Munich (UPC_CFI_402/2023 and others) holds the R.265.2(c) cost decision is mandatory of the court's own motion even absent any application; Paris CD (UPC_CFI_167/2024) holds that absent a costs request "no decision under Rule 265(2)(c) should have been issued"; Nordic-Baltic (UPC_CFI_380/2023) treats R.265's cost reference as merely "a reminder," not an independent basis. (2) The R.263.3 / R.265 boundary for partial dropping: Milan routes partial-patent dropping through R.263.3 (no immediate costs); Munich and Mannheim route partial-defendant dropping through R.265.1.
Evidence (cited decisions): - UPC_CFI_472/2024 (Milan LD) — R.263, unlike R.265, does not regulate proceeding costs because proceedings continue; partial-patent dropping is an unconditional R.263.3 limitation. - UPC_CFI_457/2023 (Düsseldorf LD) — whether an admitted restriction is also a partial R.265 withdrawal need not be decided at the time it is admitted; deferred to the final R.118.5 cost decision. - UPC_CFI_167/2024 (Paris CD) — the R.265.2(c) cost decision is not ex officio: absent a request, no cost ruling issues. - UPC_CFI_402/2023 (Munich LD) — R.265.2(c) requires a cost decision ex officio, no request needed. - UPC_CFI_380/2023 (Nordic-Baltic RD) — R.265's cost-decision requirement is a "reminder," not an independent basis. - UPC_CFI_41/2024 (Munich LD) — R.263.2 limbs (a)/(b) independent and cumulative, no discretion if either fails. - UPC_CFI_44/2024 (CoA) — R.263.2(b) requires courts to consider mitigations (extensions, separation) before refusing leave. - UPC_CFI_424/2023 (The Hague LD) — reads "reasonable diligence" generously where the amendment responds to the opponent's later partial withdrawal.
Practical impact. Litigants cannot predict whether dropping a patent or a defendant is a costless R.263.3 limitation or a cost-bearing R.265 withdrawal — the characterisation shifts who pays for the abandoned part, and divisions split on whether to decide it immediately or defer it. The R.265.2(c) split means a withdrawing party in Paris CD escapes a cost decision absent a request, but in Munich the court imposes one ex officio.
Proposed direction (signal only). Add an express cost-allocation clause to R.263 for partial limitations and a cross-reference defining the R.263.3 / R.265 boundary, and codify whether the R.265.2(c) cost decision is mandatory ex officio or request-contingent.
Confidence: strong. 20+ distinct cases across the CoA, six LDs and three CD seats. Minor caveat: a cost-percentage detail for UPC_CFI_472/2024 rests on a holding summary rather than a retrieved verbatim quote.
The rule. R.295: "The Court may stay proceedings: (a) where it is seized of an action relating to a patent which is also the subject of opposition … before the European Patent Office … where a decision in such proceedings may be expected to be given rapidly; … (d) at the joint request of the parties; … (m) in any other case where the proper administration of justice so requires."
What courts are doing. R.295 lists eleven situations in which the Court "may" stay but supplies almost no decisional standard. Three gaps recur. (1) Partial stays. R.295(d) speaks only of a stay "at the joint request of the parties" and is silent on whether a stay can be partial. Munich (UPC_CFI_815/2024) read the silence as a prohibition — "this provision does not provide for a partial stay" — and refused even to approximate one via case management. Milan (UPC_CFI_472/2024, followed by UPC_CFI_181/2025) reached the opposite result: on a joint request "may = shall" (no discretion) and a partial stay for one patent is possible. (2) The "rapidly" criterion of R.295(a), implementing Art.33(10) UPCA, is undefined; divisions converged on a relative benchmark — a decision expected "before or just after" the UPC/CoA oral hearing (UPC_CoA_237/2025 — Juul v NJOY; Hamburg UPC_CFI_851/2025 — Dyson v Dreame) — but require concrete evidence of the EPO date (Paris CD UPC_CFI_263/2023) and weigh the Preamble's one-year-to-hearing target heavily against a stay. (3) The residual R.295(m) catch-all is policed only by a court-made "yields to the more specific provision" rule (CoA UPC_COA_511/2024) and an Art.6 ECHR / Art.47 Charter "default is not to stay" gloss.
Divergence. A clean split on partial stays: Munich (UPC_CFI_815/2024) holds R.295(d) authorises only a full stay, no textual basis for a partial one; Milan (UPC_CFI_472/2024 / UPC_CFI_181/2025) holds a partial stay is possible and that a joint request makes the stay mandatory — on identical wording. Secondary divergence on the "rapidly" threshold (relative "around-the-hearing" benchmark vs. concrete near-term-date evidence).
Evidence (cited decisions): - UPC_CFI_815/2024 (Munich LD) — R.295(d) permits only a full stay; "does not provide for a partial stay." - UPC_CFI_472/2024 (Milan LD) — opposite: "may = shall" on joint request, and a partial stay for one patent is possible. - UPC_CFI_181/2025 (Milan LD) — follows UPC_CFI_472/2024. - UPC_CoA_237/2025 (CoA, Juul v NJOY) — "rapidly" is a relative benchmark anchored to the possible CoA oral-hearing date. - UPC_CFI_851/2025 (Hamburg LD, Dyson v Dreame) — a "rapid" EPO decision means one expected around the UPC oral hearing; mere acceleration insufficient. - UPC_CFI_263/2023 (Paris CD) — R.295(a) requires concrete evidence of the expected EPO date — a court-built evidentiary test. - UPC_COA_511/2024 (CoA) — residual R.295(m) unavailable where more specific provisions govern. - UPC_CFI_210/2023 (Mannheim LD) — R.295(d) joint-request requirement is strict; R.295(m) cannot compel a unilateral stay over a party with a legitimate interest in judgment.
Practical impact. Whether a litigant can obtain a partial stay depends entirely on which division is seised — Milan grants it, Munich forecloses it on identical R.295(d) wording. The undefined "rapidly" criterion makes EPO-opposition stay applications a gamble, and the strong Preamble one-year default means most stays are refused. R.295(m) provides no ex ante guidance for coordinating parallel EPO, national and FRAND proceedings.
Proposed direction (signal only). R.295 should expressly authorise (and set criteria for) partial stays under (d) and codify a structured "rapidly" test for (a)-type EPO-opposition stays anchored to the oral-hearing date and the one-year target.
Confidence: strong. 20+ decisions across 10 divisions with a clean documented partial-stay split (Munich UPC_CFI_815/2024 vs. Milan UPC_CFI_472/2024 / UPC_CFI_181/2025).
The rule. R.370.9: "Fixed and value-based fees may be reimbursed as follows: … (b) In case of the withdrawal of an action [Rule 265] the party liable for the Court fees will be reimbursed by 50% if the action is withdrawn before the closure of the written procedure. (c) If the parties have concluded their action (i) by way of settlement at the Centre …" R.360 lets the Court dispose by order of an action that "has become devoid of purpose."
What courts are doing. The fee schedule (R.370.2–.9) is keyed to a closed list of "actions" and the reimbursement rule speaks only of "withdrawal of an action [Rule 265]" or Centre settlement. Three recurring procedural events fall outside that wording, and the courts uniformly fill each by analogy, expressly labelling the omission a "planwidrige Regelungslücke." (1) PI-application withdrawal: R.370.9(b) names only "action" (cross-referencing R.265, a Klage). Munich (UPC_CFI_755/2024, UPC_CFI_515/2024), Nordic-Baltic and The Hague hold direct application is ruled out but apply the rule analogously because the incentive rationale is comparable. (2) Mooted counterclaims under R.360: UPC_CFI_300/2023 applies R.370.9(b)/(c) by analogy to a counterclaim mooted by EPO revocation. (3) FRAND-licence counterclaims, which carry no fee in the R.370.2 table: UPC_CFI_487/2023 holds the absence of a fee provision is "an unplanned gap in the Rules of Procedure." The CoA endorses the technique on appeal (UPC_CoA_611/2024; UPC_CoA_641/2025 — Sanofi/Regeneron v Amgen: R.370.9(b)/R.265 apply "by analogy to appeal withdrawals").
Divergence. A narrow but real split on PI reimbursement: Munich, Nordic-Baltic and The Hague apply R.370.9(b) to PI-application withdrawals by analogy; Mannheim and Milan (UPC_CFI_472/2024) refuse — holding the rule is limited to "actions," PI stages differ, and PI fees are already reduced, so "neither direct nor analogous application." On mootness and FRAND-counterclaim fees there is no divergence — a single consistent analogy line.
Evidence (cited decisions): - UPC_CFI_487/2023 (Munich LD) — FRAND-counterclaim fee absence is "an unplanned gap"; filled by analogy. - UPC_CFI_300/2023 (Munich LD) — R.370.9 applied by analogy to a revocation counterclaim mooted under R.360. - UPC_CFI_755/2024 (Munich LD) — "Es besteht eine Regelungslücke"; R.370.9(b) applied by analogy to PI-application withdrawal. - UPC_CFI_515/2024 (Munich LD) — direct application ruled out, analogous application appropriate because the gap is planwidrig. - UPC_CoA_611/2024 (CoA, DISH v Aylo) — endorses the analogy at appellate level (60% refund on appeal withdrawal before close of written procedure). - UPC_CoA_641/2025 (CoA, Sanofi/Regeneron v Amgen) — R.370.9(b)(i) applies by analogy to appeal withdrawals; R.370.9/.10 otherwise exhaustive. - UPC_CFI_472/2024 (Milan LD) — refuses the analogy for PI: reimbursement "is limited to actions … and does not apply to applications for provisional measures, either directly or by analogy." - UPC_CFI_249/2023 (Munich LD) — R.360 itself applied by analogy to mooted PI proceedings; "abweisen" read as "disposing of."
Practical impact. A litigant who withdraws a PI application, or whose counterclaim is mooted by parallel EPO revocation, cannot read R.370 and predict whether or how much of the court fee comes back — Munich/Nordic-Baltic/The Hague refund by analogy; Mannheim/Milan refuse outright. FRAND-counterclaim filers face an undefined fee for a procedural step the table never priced. This is forum-dependent cost exposure on procedural mechanics that should be uniform.
Proposed direction (signal only). The Administrative Committee should extend the R.370 fee table and R.370.9 reimbursement schedule to expressly cover PI-application withdrawals, R.360 mootness disposals, and FRAND/licence counterclaims, eliminating the need for divisional analogy.
Confidence: strong. Eight cases across five divisions plus two CoA decisions; the "Regelungslücke" framing and the Munich-vs-Mannheim/Milan PI split are quoted verbatim. Note: a "CoA_2/2024 fee-by-analogy" reference surfaced in the survey could not be confirmed in the database; UPC_CoA_611/2024 and UPC_CoA_641/2025 are used as the appellate anchors instead.
The rule. R.151: a cost application "shall contain … (d) an indication of the costs for which compensation is requested, which may include recovery of court fees and costs of representation, of witnesses, of experts, and other expenses; and (e) the preliminary estimate of the legal costs …" and must be lodged "within one month of service of the decision."
What courts are doing. R.151 lists what a cost application "shall contain" and fixes a one-month deadline, but says nothing about (a) interest on the sum eventually awarded, (b) whether the costs of conducting the cost proceeding are recoverable, or (c) which "decision" starts the month running where a first-instance assessment is later superseded on appeal. On interest, Munich (UPC_CFI_249/2023, UPC_CFI_696/2024) and The Hague hold the omission is a deliberate legislative choice: because interest is expressly provided for damages (R.131.2(a)) but not in R.151, there is "no unplanned legislative gap," and the remedy for delay is instead immediate provisional cost recovery under R.221.1(d)/R.242.2(a). On "no costs upon costs," the CoA (UPC_CoA_618/2024 — Hanshow v VusionGroup; UPC_CoA_153/2025 — Tiroler Rohre) and following LDs read R.150 as a "summary" track where a successful party bears its own cost-proceeding costs, with only the appeal court fee carved out — a limitation found nowhere in the text. On the deadline, Düsseldorf (UPC_CFI_1110/2025 — Seoul Viosys) holds a CoA reversal strips the first-instance assessment of its basis and allows recovery of already-paid sums in fresh post-appeal proceedings.
Divergence. Two splits. (1) Silence-as-deliberate vs. silence-as-gap: Munich and The Hague treat R.151's omission of interest as a deliberate choice barring analogy to R.125/131 — yet the same Munich decision (UPC_CFI_249/2023) treats R.151/R.118.5's silence on PI cost orders as a fillable gap to be closed by analogy. On interest itself the line is consistent, but the methodology — when silence is deliberate and when it is a lacuna — is incoherent. (2) "No costs upon costs": uniformly applied with a court-fee carve-out, but the carve-out's boundary is articulated differently (Mannheim limits it to "the court fee of an appeal"; Hanshow more broadly to "the court fee").
Evidence (cited decisions): - UPC_CFI_249/2023 (Munich LD) — no interest on cost awards; analogy to R.125/131 excluded as deliberate omission; same decision applies R.118.5 by analogy to PI cost orders, calling that a regulatory gap. - UPC_CFI_696/2024 (Munich LD) — no legal basis for interest on fixed costs; confirms the deliberate-silence reading. - UPC_CoA_618/2024 (CoA, Hanshow v VusionGroup) — successful party in R.150 cost proceedings bears its own costs; only the court fee recoverable — read into a rule that does not say so. - UPC_CoA_153/2025 (CoA, Tiroler Rohre) — withdrawal-termination counts as a R.150.1 decision triggering the same cost-assessment track. - UPC_CFI_335/2025 (Mannheim LD) — even a successful party in cost proceedings carries its own costs, the limited exception being the court fee of an appeal. - UPC_CFI_355/2023 (Düsseldorf LD) — cost proceedings are summary; "costs upon costs" not envisaged. - UPC_CFI_1110/2025 (Düsseldorf LD, Seoul Viosys) — a CoA reversal strips the first-instance assessment of its basis; already-paid amounts recoverable in post-appeal proceedings. - UPC_CoA_1/2024 (CoA) — the one-month deadline runs from service of the CoA decision; a separate application is required for appeal costs.
Practical impact. A prevailing party that waits years for assessment recovers no interest on its outlay, while a losing party has every incentive to delay — the courts' answer (provisional recovery under R.221.1(d)) only helps parties who obtained a provisional cost order. The "no costs upon costs" rule means contesting an inflated bill is economically self-defeating. And because the deadline's trigger and the effect of an appeal reversal are nowhere specified, applicants risk either forfeiting recovery by missing the one-month window or filing prematurely against a decision that is later reversed.
Proposed direction (signal only). Amend R.151/R.152 to state expressly (i) whether statutory interest accrues on awarded costs and from when, (ii) whether and to what extent the costs of the cost proceeding are recoverable, and (iii) the deadline's trigger where an appeal supersedes a first-instance assessment.
Confidence: strong on the underlying holdings (two reasoned Munich LD decisions, two CoA decisions, Düsseldorf UPC_CFI_1110/2025). Weaker point flagged: the "internal inconsistency" characterisation of UPC_CFI_249/2023 is this memo's synthesis of two holdings within one decision, not an express judicial statement of contradiction.
The rule. R.19.1: "Within one month of service of the Statement of claim, the defendant may lodge a Preliminary objection concerning: … (c) the language of the Statement of claim [Rule 14]." R.323.2: "The President shall invite the other party to indicate, within 10 days, its position …" — followed by R.323.3 empowering the President, having consulted the panel, to order the language change. The rule names the applicant's Application and the respondent's position statement, but no applicant reply.
What courts are doing. Courts read R.19(c) narrowly: by cross-referring only to R.14, the language preliminary objection reaches solely compliance with Art.49(1)–(2) UPCA, not the broader Art.49(5) fairness/convenience question. Paris (UPC_CFI_583/2024) and the Brussels CFI President (UPC_CFI_871/2026 — Esko v In(k)control) hold an Art.49(5)/R.323 fairness application is simply not a R.19(c) preliminary objection — so it escapes R.19's one-month deadline and formalities, but also gets no preliminary-objection track. The R.19.1 grounds list is treated as exhaustive (CoA UPC_CoA_188/2024; Munich UPC_CFI_15/2023; Munich CD UPC-CFI-2296/2025): stay, security-for-costs, abuse, standing and res judicata are all excluded. On R.323's mechanics, the central gap is the absence of an applicant reply: the CoA (UPC_CoA_354/2024 — Apple) held squarely that "in R.323 RoP there is no reply to a respondent's statement under R.323.2 RoP," and that further pleadings need leave; the President may simply ignore an unsolicited reply. Courts also had to invent the timing rule themselves — R.323.1's "shall include in the Statement of Defence" is read as a deadline, not a fixed filing moment (UPC_CFI_239/2023; UPC_CoA_207/2024).
Divergence. No clear divergence on outcome — a consistent CoA-led line is followed by the CFI President and divisions. The only nuance is the doctrinal label for where the fairness application lives (Art.49(4) vs. Art.49(5)) — a labelling inconsistency the rule text invites by not naming its own statutory base in R.323.1.
Evidence (cited decisions): - UPC_CFI_583/2024 (Paris LD) — R.19(c) narrows language objections to Art.49(1)–(2) compliance; no remedy for convenience/fairness challenges. - UPC_CoA_354/2024 (CoA, Apple) — R.323 provides no reply to the respondent's R.323.2 statement; further pleadings require leave. - UPC_CFI_871/2026 (CFI President, Brussels LD, Esko v In(k)control) — R.323.2 sets the only foreseen submissions; further unsolicited submissions "should not be considered." - UPC_CoA_207/2024 (CoA) — R.323 does not mandate filing the application in the Statement of Defence; timing wording treated as a non-binding outer limit. - UPC_CFI_239/2023 (CFI President) — R.323.1's "shall include in the Statement of Defence" construed as a deadline, not a waiting period. - UPC_CFI_15/2023 (Munich LD) — R.19.1 grounds are exhaustive; stay, TQJ allocation and security-for-costs fall outside. - UPC_CoA_188/2024 (CoA) — preliminary-objection grounds are an exhaustive list; abuse and manifest unfoundedness are not preliminary objections. - UPC-CFI-2296/2025 (Munich CD) — confirms exhaustive enumeration; standing and res judicata excluded.
Practical impact. A defendant seeking the patent language must use the R.323 route, not the R.19(c) objection — and once the respondent files its R.323.2 position, the applicant has no right of reply and the President can decide (and ignore an unsolicited reply) without hearing the rebuttal. This makes the single 10-day respondent statement decisive, pressuring applicants to front-load everything into the original Application. The narrow R.19(c) scope means convenience/fairness disputes never get early, appealable preliminary-objection treatment.
Proposed direction (signal only). R.323 should expressly grant the applicant a short reply (or codify the leave standard for further submissions) and state its filing window as an explicit deadline; R.19(c) could be clarified to confirm it covers only Art.49(1)–(2) compliance and to signpost R.323 for fairness challenges.
Confidence: strong. Multiple on-point CoA decisions (UPC_CoA_354/2024, UPC_CoA_207/2024) plus CFI President and divisional rulings, and a consistent exhaustive-list line on R.19.
The rule. R.220.3: "In the event of a refusal of the Court of First Instance to grant leave within 15 days of the order of one of its panels a request for a discretionary review to the Court of Appeal may be made within 15 calendar days … Rule 333.3 shall apply mutatis mutandis. The request shall set out the matters referred to in Rule 221.2." — i.e. the rule names the remedy and the deadline but states no substantive criterion. R.333.1: case-management orders "shall be reviewed by the panel, on a reasoned Application by a party."
What courts are doing. The only articulation of a standard lives in Preamble point 8 ("necessary to ensure consistent application and interpretation of the RoP") — which is not part of any operative rule. The Court of Appeal has therefore judge-made a two-pronged gateway: the impugned order must be (i) manifestly erroneous, or (ii) raise a fundamental question of law whose review is necessary for the consistent application/interpretation of the RoP (UPC-CoA-74/2026 — ASC v Motorola; UPC-COA-0000056/2026). Repeated CoA orders stress that alleging mere incorrectness is insufficient and that new arguments not put to the CFI are barred (UPC_CoA_489/2024 — Motorola; UPC_CoA_570/2024). On the R.333 side, "case-management decision or order" is read broadly by most LDs but the panel's competence boundary is contested — particularly over R.262.1(b) third-party access.
Divergence. Two splits. (1) Structure of the gateway: UPC-CoA-67/2026 (standing judge) reads "manifest error AND fundamental question" as cumulative; UPC-COA-0000056/2026 and UPC-CoA-35/2026 read them as alternatives. (2) Whether a pure interpretative/consistency justification suffices: UPC-COA-0000052/2026 grants review on the consistency prong alone (divergent LD positions on R.333/R.262.1(b)), whereas UPC_CoA_469/2024 holds the interpretative question alone insufficient absent case-specific misuse of discretion. On R.333 scope, Munich (UPC_CFI_1234/2025) denies panel competence over a JR's R.262.1(b) access decision, while Mannheim (UPC_CFI_471/2023) treats R.262A/access orders as squarely reviewable.
Evidence (cited decisions): - UPC-CoA-74/2026 (CoA, ASC v Motorola) — states the two-pronged standard; ASC's listing of reasons the order is incorrect, plus new arguments, fails because it never argues the errors are manifest. - UPC-CoA-67/2026 (CoA, standing judge) — reads the gateway as cumulative ("requires both manifest error and a fundamental question"). - UPC-COA-0000056/2026 (CoA) — reads the gateway as alternative ("manifest incorrectness OR a fundamental question"). - UPC_CoA_489/2024 (CoA, Motorola) — discretionary review is not a general second instance; a systemic justification is required. - UPC_CoA_570/2024 (CoA) — applicant must affirmatively demonstrate review is necessary for consistent application; mere disagreement insufficient. - UPC_CoA_469/2024 (CoA) — an interpretative question alone is insufficient without case-specific misuse of discretion — a stricter line. - UPC-COA-0000052/2026 (CoA) — grants review purely on the consistency prong where LDs diverge on R.333/R.262.1(b). - UPC_CFI_1234/2025 (Munich LD) — panel review inadmissible against a JR's R.262.1(b) decision (no competence) — narrowing R.333. - UPC_CFI_471/2023 (Mannheim LD) — "case-management decision" read broadly to cover R.262A/access orders — contrasting Munich.
Practical impact. Litigants seeking to appeal non-final and case-management orders face genuine unpredictability: the operative rule gives no standard, so whether leave is granted turns on which unwritten formulation a particular standing judge applies and how generously the "consistency" justification is read. A party that frames its request around mere incorrectness — the intuitive appellate posture — will almost always lose, and arguments not raised below are forfeited. The fragmented R.333 scope (especially over access orders) means parties cannot reliably predict whether the route to challenge a JR order is panel review, direct appeal, or discretionary review, risking inadmissibility on the wrong procedural track.
Proposed direction (signal only). Codify the discretionary-review standard — and whether the prongs are cumulative or alternative — in the text of R.220.3 itself rather than leaving it to be inferred from Preamble point 8, and clarify the R.333 panel-competence boundary for JR access decisions.
Confidence: strong. 10+ CoA/CFI decisions directly on R.220.3/R.333, including the Motorola line and ASC v Motorola; the cumulative-vs-alternative split is evidenced by verbatim quotes. Note: some 2026 numbers are stored in normalised form (e.g. UPC-COA-0000052/2026).
Four patterns run across the eleven clusters and are themselves diagnostic of structural weaknesses in the RoP, independent of any single rule.
A. The RoP delegates the test but not the criteria. The Court's framework character is a deliberate design choice, but the recurring failure mode is a rule that grants a discretion ("the Court may …") without naming the factors that structure it. R.158 ("adequate security"), R.190 (production "on a reasoned request"), R.211.2 ("sufficient degree of certainty"), R.262 ("reasoned request" / "legitimate reasons"), R.295 ("rapidly" / "proper administration of justice"), R.9.3 (extension "even retrospectively"), and R.220.3 (discretionary review) all share this shape. In every case the divisions have supplied the missing criteria — competently, but independently, which is how divergence arises.
B. "Planwidrige Regelungslücke" as a recurring judicial verdict. The courts repeatedly diagnose unplanned gaps and fill them by analogy: R.370 fee refunds (PI withdrawal, mooted counterclaims, FRAND counterclaims), R.158 release/revocation (borrowing R.352.2 / R.335), R.211/R.360 mootness, R.151/R.118.5 PI cost orders. That the courts must reach for analogy this often is itself a signal that the fee schedule and several procedural mechanisms were drafted against an incomplete enumeration of the procedural events that actually occur. Notably, the courts do not even agree on when silence is a fillable gap and when it is a deliberate legislative choice (R.151 interest is the sharpest example) — so the gap-filling methodology is itself unsettled.
C. The confidentiality/access complex is the single largest pressure point. R.262/R.262A is litigated more than twice as often as any other rule, and it radiates into others: production under R.190 routes confidentiality through R.262A; the R.220.3/R.333 review split is largely about who can review a R.262.1(b) access decision. A focused codification of the access threshold, the balancing test, and club composition would reduce litigation across three clusters at once.
D. Appellate guidance lags first-instance need, and the gatekeeper for that guidance is itself undefined. On the issues where the CoA has spoken (R.158 criteria, R.211 standard of proof, R.262 reasoned request), first-instance practice has largely converged. The persistent divergence sits where the CoA has not yet ruled (auxiliary requests in PI — expressly deferred in UPC_CoA_382/2024; partial stays under R.295(d)) or where the route to the CoA is itself the contested question (R.220.3). The discretionary-review gateway that is supposed to deliver harmonisation is the one rule with no standard in its operative text — a structural bottleneck on the Court's own ability to self-correct.
Prepared by Lexy (UPC patent-litigation analysis), yoUPC.org. Evidence base: ~1,950 classified UPC decisions in data.ai_judgment_notes; rule text from the youpc-laws UPCRoP corpus. Every cited case number was verified to exist in the database. This is procedural analysis for a rules-improvement consultation, not legal advice.
Eleven proposals derived from Phase 1's weakness analysis, formatted for direct submission to the UPC RoP Revision form on fdbck.msbls.de (slug Lj8qVGQeuDnSF1BPWEHo8Y92u8QIrP8o). Each follows the form's required 4-part structure. Submit each as a separate response.
Form intro (verbatim): "Proposals should be specific and actionable. Use clear legal language, follow the 4 part structure set out below." Field 2 uses mark-up mode —
~~deletions~~in strikethrough and insertions in bold; wholly new paragraphs/rules are rendered in full and bolded as insertions. Wiki-style rule links in the source corpus are rendered here as plain "Rule N".
R.262.1(b), R.262.6 (amendment); new R.262.4(d) (insertion); R.262A.6 (amendment); new R.262A.6a and new R.262A.6b (insertion).
R.262.1(b) (amended):
(b) written pleadings and evidence, lodged at the Court and recorded by the Registry, in so far as they relate to the substance of a party's case, shall be available to the public upon reasoned request to the Registry, save that registry correspondence, formality checks and communications with the European Patent Office shall not fall within this paragraph; the decision is taken by the judge-rapporteur after consulting the parties.
R.262.4 (new sub-letter (d) inserted; (a)–(c) unchanged):
[R.262.4(a)–(c) unchanged] (d) a concrete, verifiable and legitimate reason why access to the information is necessary, it being insufficient to invoke a generic or unverifiable purpose; and the applicant shall be represented in accordance with Rule 8.
R.262.6 (amended):
R.262A.6 (amended):
R.262A.6a (new paragraph inserted):
6a. In determining the number and identity of the persons referred to in paragraph 1, and whether a particular nominated person shall be granted access, the Court shall have regard in particular to: (a) the role of that person in the proceedings and whether access is required for the performance of that role; (b) the relevance of the information or evidence to that role; and (c) the trustworthiness of that person with respect to the confidential information. The status of a nominated person as an employee of a party shall not of itself be a ground for excluding that person.
R.262A.6b (new paragraph inserted):
6b. Where the applicant seeks redaction of, or restricted access to, particular information, the grounds under paragraph 2 shall be substantiated in respect of each item or category of information for which protection is sought.
R.262 grants the public access to pleadings and evidence "upon reasoned request," and R.262A lets the Court restrict access to confidential material, but neither rule defines what a reasoned request must contain, the access threshold, or how a confidentiality club is sized — so the divisions invented all three, producing both unwritten requirements and a genuine split.
On the public-access content and threshold, the Munich Central Division (UPC_CFI_1/2023; UPC_CFI_75/2023, Astellas) requires a "concrete, verifiable, legitimate reason" stating documents, purpose and necessity, and rejects generic purposes such as education or training; the Court of Appeal canonised this in UPC_CoA_404/2023 (Ocado v AutoStore) and added an unwritten requirement that the applicant be represented. The Nordic-Baltic RD (UPC_CFI_8/2023) applies a markedly lower presumption-of-access "credible explanation" bar. I codify the stricter Munich/Court-of-Appeal line because the Court of Appeal sits above the divisions and has settled it. The scope limitation — substantive pleadings only, excluding registry/EPO correspondence — is settled in UPC_CFI_1/2023 and UPC_CFI_14/2023 and is codified verbatim.
On confidentiality clubs, the Court of Appeal in UPC_CoA_621/2024 (Daedalus Prime v Xiaomi) fixed the three-factor sizing test (role, relevance, trustworthiness) and held employee status alone cannot exclude a nominee; Hamburg (UPC_CFI_169/2024) declared a UPC-wide minimum barring attorneys'-eyes-only clubs; Düsseldorf (UPC_CFI_355/2023) requires per-redaction substantiation. These are codified in new R.262A.6a/6b. The amendment makes settled judicial consensus visible without creating new doctrine.
The reasoned-request criteria go into R.262.4 as a new sub-letter (d) because R.262.4 already enumerates the Application's required content (a)–(c); adding (d) keeps the content requirement operative where a reader already looks for it, rather than burying it in the discretionary balancing of R.262.6. The representation requirement is anchored to Rule 8 (general representation rule) so it inherits existing definitions. R.262.6 is amended only to make paragraph 4 a gateway, preserving the existing confidentiality-balancing clause untouched.
The "concrete, verifiable and legitimate reason" wording is the Court of Appeal's own phrase from UPC_CoA_404/2023, not a new standard; "it being insufficient to invoke a generic or unverifiable purpose" forecloses the Nordic-Baltic credible-explanation reading without barring legitimate case-law-monitoring once it is made concrete — it raises the quality of the reason, not its subject-matter.
The club criteria go into R.262A.6a immediately after the floor in R.262A.6, keeping all club-sizing rules contiguous. One objection: codifying "no attorneys'-eyes-only" may seem to remove flexibility. But the floor already mandates "at least one natural person from each party"; the added sentence merely states the necessary corollary the Court of Appeal and Hamburg already enforce, making it visible rather than discoverable only through litigation. No existing paragraphs are renumbered; 6a/6b are inserted as lettered sub-paragraphs.
R.158.1 (amendment); new R.158.1a, new R.158.1b, new R.158.1c (insertions); new R.158.6 (insertion). Existing R.158.2–R.158.5 unchanged and not renumbered.
1. At any time during proceedings, following a reasoned request by a defendant, the Court may order the claimant to provide, within a specified time period, adequate security for the legal costs and other expenses incurred and/or to be incurred by the requesting party, which the claimant may be liable to bear. No security may be ordered against a defendant, and the relative financial position of the parties shall not of itself be a ground for ordering security. Where the Court decides to order such security, it shall decide whether it is appropriate to order the security by deposit or bank guarantee.
1a. The Court may order security only where the requesting party shows, on specific facts, a legitimate and real concern that a future order for costs against the other party would not be recoverable, or would be enforceable only in an unduly burdensome way. The mere fact that the other party is domiciled or has its assets outside the territory of the Contracting Member States shall not, of itself, satisfy this requirement.
1b. The amount of security shall be proportionate to the costs which the requesting party may reasonably expect to recover, having regard to the ceiling for recoverable costs set under Rule 152.2. The applicable ceiling shall serve as a reference point and not as an automatic amount; the Court may set a lower amount where the full ceiling would exceed the costs reasonably to be expected.
1c. Adequate security shall be provided by deposit or by a bank guarantee enforceable in a Contracting Member State. Insurance for adverse costs and guarantees not so enforceable shall not, of themselves, constitute adequate security.
[R.158.2 unchanged] [R.158.3 unchanged] [R.158.4 unchanged] [R.158.5 unchanged]
6. The Court may, following a reasoned request by either party, vary or revoke an order made under paragraph 1 where there has been a material change in the circumstances on which the order was based. Rule 158.2 shall apply. Security provided shall be released when it is no longer required, in particular when the proceedings are concluded and any liability for costs has been discharged.
R.158 lets the Court order "adequate security" for costs on a "reasoned request" but specifies no test, no amount, no accepted form, and no mechanism to release, reduce or revoke — so each division has filled the gaps differently, making security a forum-shopping vector. The Court of Appeal has now settled the core points, and this amendment makes that settled law visible rather than leaving it to analogy. New R.158.1a codifies the two-factor "legitimate and real concern" recoverability/enforceability test established in UPC_CFI_252/2023 (NanoString v Harvard) and elevated by the CoA in the Audi v Network System Technologies line (UPC_CoA_217/2024). The amended R.158.1 codifies UPC_CoA_393/2025 (AorticLab v Emboline) and UPC_CFI_525/2024 (Visibly v Easee) — security runs only against a claimant — and UPC-COA-0000001/2026 (Adobe v KEEEX), holding relative financial position is not a criterion. On amount, the divisions split: Lisbon (UPC_CFI_850/2026 — Transsion v Ericsson) and Paris CD (UPC_CFI_164/2024 — Microsoft v Suinno) anchor at the R.152 ceiling, while Düsseldorf (UPC_CFI_100/2024 — Ona Patents v Google) rejects the ceiling as an automatic anchor. New R.158.1b adopts Düsseldorf's reading — ceiling as reference, not automatic amount — as the more proportionate rule, consistent with R.152's "reasonable and proportionate" standard. New R.158.1c codifies the restrictive form reading of UPC_CoA_301/2024 (ICPillar v ARM) and UPC_CFI_617/2025 (Munich), rejecting ATE insurance and non-EU guarantees. New R.158.6 gives the express vary/revoke/release power the courts now graft on by analogy or as an "implied power" (UPC_CFI_164/2024; UPC_CFI_252/2023 releasing via R.352.2).
The test, quantum, form and release rules are inserted as new sub-paragraphs 1a–1c immediately after R.158.1 because they qualify the single substantive power that paragraph 1 confers; placing them there keeps them operative and reviewable rather than relegating them to a preamble. Existing paragraphs 2–5 are left untouched and unrenumbered — the lettered "1a/1b/1c" form avoids cascading cross-reference breakage in R.158.2–5 and elsewhere in the RoP. Paragraph 1 itself is rewritten from "one party"/"other party" to "defendant"/"claimant" because the CoA has confirmed security runs only one way; retaining the neutral wording would preserve the false symmetry the rule superficially offered. The quantum wording "proportionate to the costs which the requesting party may reasonably expect to recover … the applicable ceiling shall serve as a reference point and not as an automatic amount" deliberately imports no fixed percentage: it neither blesses the 50%-of-ceiling anchor nor forbids it, leaving the Court to calibrate within the R.152 ceiling. One objection is that R.158.1c's enforceability requirement could exclude a solvent foreign guarantor; but the requirement attaches to enforceability within a Contracting Member State, mirroring the CoA's concern in UPC_CoA_301/2024 that security be realisable without the very recovery difficulty it is meant to cure, and R.158.6 lets the Court vary the order if circumstances change.
R.190.1 (amendment); new R.190.1a (insertion). Existing R.190.2–7 unchanged and not renumbered.
1. Where a party has presented reasonably available and plausible evidence in support of its claims and has, in substantiating those claims, specified evidence which lies in the control of the other party or a third party, the Court may on a reasoned request by the party specifying such evidence, order that other party or third party to produce such evidence. The applicant need not prove the relevance of the specified evidence conclusively; it shall be sufficient that the evidence is plausibly relevant to a fact in dispute. For the protection of confidential information the Court may order that the evidence be disclosed to certain named persons only and be subject to appropriate terms of non-disclosure, in accordance with Rule 262A.
1a. In deciding upon a request under paragraph 1, the Court shall be guided by the principles of necessity, relevance and proportionality. The Court shall in particular have regard to: (a) whether the requested evidence is specified with sufficient precision; a request shall not be granted in so far as it seeks categories of evidence described only in general terms, or seeks all documents of a given kind, such that the request amounts to a search for evidence whose existence or content is merely speculated; (b) the importance of the requested evidence for deciding the issue in question, weighed against the burden, cost and confidentiality interests that production would impose on the producing party or third party; and (c) whether the request is directed to substantiating a claim or defence already advanced, rather than to identifying a new claim or theory not otherwise pleaded.
[R.190.2 unchanged] [R.190.3 unchanged] [R.190.4 unchanged] [R.190.5 unchanged] [R.190.6 unchanged] [R.190.7 unchanged]
Rule 190 lets the Court order production of specified evidence in a party's control, but its text fixes no threshold for how relevant or how specific the request must be, nor any proportionality criterion — so divisions have built the operative test entirely in case law. Munich expressly found the gap: UPC_CFI_208/2023 held "the RoP contains no rule" requiring conclusive proof of relevance before production is ordered. To fill it, The Hague (UPC_CFI_327/2024) and Brussels (UPC_CFI_1357/2025, GC Aesthetics v Establishment Labs) constructed a "four cumulative conditions" framework with two-stage proportionality, none of which appears in the rule. To police the missing specificity threshold, Mannheim (UPC_CFI_216/2023) and Munich (UPC_CFI_450/2024) invented the "fishing expedition"/Ausforschung doctrine, rejecting requests for "all" licences or requests identical across parallel proceedings and placing the burden on the applicant. The Court of Appeal (UPC_CoA_298/2024) then imported a "strict necessity, relevance and proportionality" standard by analogy to EU Directives 2004/48 and 2014/104, confirming the rule supplies no native test. Munich (UPC_CFI_65/2024) added a purpose limit: R.190 cannot be used to find a new infringement theory after the first proves untenable. This proposal codifies that settled, convergent judicial consensus rather than any novel doctrine: the CoA's necessity/relevance/proportionality standard, the anti-fishing specificity requirement, and the purpose limit. There is no clean doctrinal split to resolve; the divisions converge, so the amendment simply makes the shared test visible and reviewable in the rule's own words.
The relevance clarification sits in paragraph 1 because it qualifies the existing "plausible evidence" ingredient already there, directly answering Munich's gap-finding in UPC_CFI_208/2023 that no conclusive proof is required. The new criteria go into a fresh paragraph 1a rather than into 1 to keep the operative grant clean and to give the structured test its own home, and 1a is inserted before the existing 2 so no later paragraph is renumbered and no cross-reference elsewhere in the RoP breaks. "Necessity, relevance and proportionality" mirrors the Court of Appeal's own formula in UPC_CoA_298/2024, so the rule adopts the CoA's words rather than coining new ones. Sub-(a) codifies the anti-fishing doctrine in functional terms — "categories described only in general terms" / "all documents of a given kind" — to capture the Mannheim and Munich case law without freezing it to particular fact patterns. Counter-reading: one might object that "plausibly relevant" lowers the bar to anything arguable. It does not — sub-(a) and the proportionality weighing in sub-(b) constrain granting, so a plausibly-relevant but unspecified or disproportionate request still fails; the two operate together, exactly as the divisions apply them. Sub-(c) encodes the UPC_CFI_65/2024 purpose limit without barring late-designated evidence, consistent with UPC_CFI_550/2024 that production power applies at any stage. The R.262A cross-reference makes the confidentiality interplay explicit, which UPC_CFI_208/2023 flagged as undefined.
R.211.2 (amendment); new R.211.2a (insertion); R.211.4 (amendment); new R.211.7 (insertion). R.30 untouched (the insertion in R.211 deliberately keeps auxiliary requests out of R.30's main-proceedings scope).
R.211.2 (amended):
New R.211.2a (inserted):
2a. The Court shall assess the validity of the patent on the basis of the patent as granted or, where applicable, as previously limited or amended with effect erga omnes. The applicant may not file auxiliary requests to amend the patent in proceedings for provisional measures, save that the Court may, exceptionally and on a reasoned request, admit a limited number of auxiliary requests where this is compatible with the summary and provisional character of the proceedings and does not impair the expedition required by the imminence relied upon. Rule 30 shall not apply by analogy to proceedings for provisional measures.
R.211.4 (amended):
New R.211.7 (inserted):
7. Where the subject-matter of an Application for provisional measures ceases to exist before the Court has decided, Rule 360 shall apply by analogy.
Rule 211 is skeletal: it requires "a sufficient degree of certainty" but fixes no quantum, names "unreasonable delay" but no start date, and is silent on both auxiliary requests and mootness. Divisions have judicially completed it, and on the proof standard and urgency clock the case law is now settled enough to codify. On the standard, the Court of Appeal in UPC_CoA_1/2024 defined "sufficient degree of certainty" as "more likely than not, on the balance of probabilities", applied in UPC_CoA_297/2024 to lift a PI — the proposed R.211.2 sentence simply restates the CoA's formula. On the urgency clock, UPC_CFI_1927/2025 (Düsseldorf) runs delay from when the applicant knew or ought to have known enough to file a promising application, and UPC_CFI_317/2024 (Lisbon) already requires the applicant to state its awareness date; R.211.4 codifies both. On mootness, R.211 is silent and courts apply Rule 360 by analogy (UPC_CFI_130/2024, The Hague) — R.211.7 makes that express. On auxiliary requests the case law is split: Lisbon (UPC_CFI_317/2024), Milan (UPC_CFI_380/2024) and Munich (UPC_CFI_693/2025) flatly reject Rule 30 by analogy as incompatible with PI's summary character, while the Court of Appeal (UPC_CoA_382/2024) expressly left the question open. I codify the dominant first-instance position — no auxiliary requests as of right — but preserve the CoA's discretion through a narrow exceptional gateway, so R.211.2a does not foreclose the line the CoA is still developing.
The proof-standard sentence is appended to R.211.2 itself, not placed in a new paragraph, because it defines a term ("sufficient degree of certainty") already used there — it operates as a legislative gloss exactly where the term appears. The wording "more likely than not, on the balance of probabilities" is taken verbatim from UPC_CoA_1/2024 so the rule codifies, rather than reinterprets, the CoA's holding. Counter-reading pre-empted: this does not raise the threshold to "clear and convincing" — it states the very standard the CoA already applies. The auxiliary-request provision sits in a new R.211.2a, adjacent to the validity sentence it qualifies, and expressly disapplies Rule 30 by analogy rather than amending Rule 30, keeping Rule 30 confined to main proceedings as its text intends. The "exceptionally … on a reasoned request" gateway is deliberate: a flat prohibition would contradict the open CoA line (UPC_CoA_382/2024), so the rule encodes the first-instance default while leaving the CoA room. Objection anticipated: "exceptionally" is vague — but it is bounded by the stated criteria (summary character, expedition, imminence), which are reviewable on appeal, unlike the present silence. The urgency clock is placed in R.211.4 beside "unreasonable delay" and ties the start to the knowledge/diligence date plus a pleading duty, mirroring the divisions. Inserting R.211.2a between paragraphs 2 and 3 avoids renumbering 3–6; the lettered "2a" preserves existing cross-references to R.211.5 and R.211.6.
R.9.2 (amendment); R.9.3(a) (amendment); new R.9.3a (insertion, between R.9.3 and existing R.9.4). Existing R.9.4 is not renumbered.
2. The Court may disregard any step, fact, evidence or argument which a party has not taken or submitted in accordance with a time limit set by the Court or these Rules. A decision to disregard material under this paragraph shall be taken by the panel and not by the judge-rapporteur acting alone by way of case management.
3. Subject to paragraph 4, on a reasoned request by a party, the Court may:
(a) extend, even retrospectively, a time period referred to in these Rules or imposed by the Court, where the requesting party shows that timely compliance was objectively impossible or excessively difficult through no fault of its own; such an extension shall be limited to the additional time strictly necessary; and
(b) shorten any such time period.
3a. A request to extend a time period retrospectively under paragraph 3(a) shall be lodged as soon as the difficulty preventing timely compliance becomes apparent and, in any event, no later than one month after expiry of the time period concerned. The Court may admit a request lodged after that period only in exceptional circumstances. This paragraph shall not apply to an extension agreed between all parties.
[R.9.1 unchanged]
[R.9.4 unchanged: "The Court shall not extend the time periods referred to in Rules 198.1, 213.1 and 224.1."]
R.9.3(a) lets the Court extend a deadline "even retrospectively" on a "reasoned request," but states no criterion for granting relief and no deadline by which a retrospective request must be lodged; R.9.2 names no deciding body. The result is division-dependent and, on the retrospective-request point, openly split. A near-uniform CFI line grafts a test onto the silence: Paris CD (UPC_CFI_412/2023) invents "objective impossibility / extreme difficulty not attributable to the party" plus a duty to apply "as soon as the difficulty is apparent"; Paris again (UPC_CFI_189/2024) confines R.9.3(a) to "justified exceptional cases"; Düsseldorf (UPC_CFI_355/2023) requires attempted timely filing plus prompt upload; The Hague (UPC_CFI_1791/2025) calls it a "narrow discretionary power" exercised "with caution." Against this, the CoA reads the textual silence the other way: UPC-COA-0000037/2026 holds the absence of any time limit positively permits a request filed after the deadline, discretion "solely for the court," and UPC-CoA-61/2026 extends that to orders and requests both lodged post-expiry. This proposal codifies the CFI substantive test (which the CoA has not disavowed — it ruled only on the missing time bar, not on the threshold) while resolving the timing split by supplying the outer limit the CoA found absent: relief stays available, but reliance on a missed deadline is no longer defeated indefinitely. The consensual carve-out preserves Munich's softer line (UPC_CFI_127/2024, consensual extensions granted "as a rule"). New R.9.2 second sentence codifies UPC_CFI_869/2025 (Düsseldorf): disregard of late material is substantive and not for the judge-rapporteur alone.
The substantive threshold is placed inside R.9.3(a) itself, not the Preamble, so it is operative and reviewable on appeal rather than mere guidance. The words "objectively impossible or excessively difficult through no fault of its own" track the Paris/The Hague formula verbatim, and "limited to the additional time strictly necessary" codifies the minimum-extension principle the divisions already apply. The new R.9.3a is a separate paragraph, not a clause appended to (a), because it governs the procedure for lodging the request (timing) rather than the merits — keeping the threshold and the time bar conceptually distinct. The one-month outer limit answers the CoA's complaint that the text supplies none, while the "exceptional circumstances" escape valve preserves the CoA's concern that genuinely late-discovered impossibility not be foreclosed; it does not reopen the indefinite window the CoA read in, because the burden shifts to the late applicant. Counter-reading: one might object that fixing a number usurps judicial discretion. It does not — discretion survives on the merits (the threshold) and on admitting truly exceptional late requests; only the open-ended default is removed. The consensual carve-out in R.9.3a's final sentence prevents the limit from disturbing agreed extensions. R.9.4 is untouched and not renumbered; R.9.3a sits before it numerically and operationally.
R.263.2(b) (amendment); new R.263.5 (insertion); R.265.2(c) (amendment).
Rule 263 — Leave to change claim or amend case
[R.263.1 unchanged]
Subject to paragraph 3, leave shall not be granted if, all circumstances considered, the party seeking the amendment cannot satisfy the Court that: (a) the amendment in question could not have been made with reasonable diligence at an earlier stage; and (b) the amendment will not unreasonably hinder the other party in the conduct of its action, the Court having regard, before refusing leave, to whether any such hindrance can be mitigated by an extension of a time period, the separation of proceedings or another case management measure.
Leave to limit a claim in an action unconditionally shall always be granted.
The Court may re-consider fees already paid in the light of an amendment.
5. Where leave is granted to limit a claim under paragraph 3, or an amendment under this Rule abandons part of an action, the proceedings continue and no cost decision in respect of the abandoned part shall issue at that time; the costs of the abandoned part shall be dealt with in the decision on costs under Rule 118.5. The complete withdrawal of an action, or the withdrawal of the action against a defendant in its entirety, is governed by Rule 265 and not by this Rule.
Rule 265 — Withdrawal
[R.265.1 unchanged]
(a) give a decision declaring the proceedings closed;
(b) order the decision to be entered on the register; and
(c) of its own motion, issue a cost decision in accordance with Part 1, Chapter 5.
[closing words of R.265.2 unchanged]
Two rules govern the commonest move in litigation — dropping part of a case — but draw the line and its cost consequence differently, and neither says where a partial limitation (R.263) ends and a partial withdrawal (R.265) begins. R.265.2(c) mandates a cost decision on withdrawal; R.263 contains no equivalent clause. This produces two splits.
First, the R.263/R.265 boundary. Milan (UPC_CFI_472/2024) reads R.263.3 expansively to cover dropping one of several patents and notes that, unlike R.265, R.263 "does not provide the regulation of the proceeding costs, because the proceedings continue" — leaving costs to the final decision. Düsseldorf (UPC_CFI_457/2023) defers characterisation to the R.118.5 cost decision. New R.263.5 codifies exactly this consensus: partial abandonment continues under R.263 with costs deferred; complete or per-defendant withdrawal falls under R.265.
Second, whether the R.265.2(c) cost decision is ex officio. Munich (UPC_CFI_402/2023) holds it mandatory of the court's own motion, even absent any application; Paris CD (UPC_CFI_167/2024) holds that absent a costs request "no decision under Rule 265(2)(c) should have been issued"; Nordic-Baltic (UPC_CFI_380/2023) treats it as a mere "reminder." I codify the Munich ex-officio position — the text already says "the Court shall," which is mandatory.
On the leave test, the CoA (UPC_CFI_44/2024) requires courts to weigh mitigations before refusing under limb (b); The Hague (UPC_CFI_424/2023) reads "reasonable diligence" generously. The amended R.263.2(b) writes the CoA's mitigation duty into the rule, against Munich's stricter, discretion-free reading (UPC_CFI_41/2024; cf. UPC_CFI_252/2023's bright-line cut-offs).
New R.263.5 is placed in R.263, not R.265, because the operative trigger is the grant of leave to limit — the rule the litigant invokes when dropping part of a case. It reproduces Milan's reasoning verbatim in substance ("the proceedings continue") so the cost-deferral is anchored to the doctrinal reason, not an arbitrary choice. The second sentence supplies the missing boundary marker, routing only complete or per-defendant withdrawal to R.265, which mirrors how Munich and Mannheim treat partial-defendant dropping. Counter-reading: one might object that "abandons part of an action" is vague — but it is deliberately functional, and the explicit carve-out for whole-action and per-defendant withdrawal removes the only cases where the line actually bites.
In R.265.2(c), "of its own motion" resolves the split in favour of Munich. One could object this overrides Paris CD's request-based reading; that is the intended effect — the chapeau already commands "the Court shall," so making it ex officio merely removes the textual ambiguity Nordic-Baltic exploited to downgrade the clause to a "reminder."
In R.263.2(b), the mitigation clause is added to limb (b) rather than as a free-standing paragraph so it operates as a condition on refusal, exactly as the CoA framed it in UPC_CFI_44/2024 — making the duty reviewable rather than hortatory. Existing paragraphs are not renumbered; R.263.5 is a clean append.
R.295(a), R.295(d), R.295(m) (amendment); new R.295.2 (insertion, structured "rapidly" criteria for stay pending EPO opposition); new R.295.3 (insertion, partial stay on joint request); new R.295.4 (insertion, subsidiarity of the (m) catch-all). The existing single un-numbered list becomes R.295.1; sub-letters (a)-(m) are retained unchanged in lettering.
295.1 The Court may stay proceedings: (the existing un-numbered chapeau is retained unchanged and designated paragraph 1)
(a) where it is seized of an action relating to a patent which is also the subject of opposition proceedings or limitation proceedings (including subsequent appeal proceedings) before the European Patent Office or a national authority where a decision in such proceedings may be expected to be given rapidly within the meaning of paragraph 2;
[R.295.1(b)-(c) unchanged]
(d) at the joint request of the parties, in accordance with paragraph 3;
[R.295.1(e)-(l) unchanged]
(m) in any other case where the proper administration of justice so requires, subject to paragraph 4.
295.2 For the purposes of paragraph 1(a), a decision may be expected to be given rapidly only where the party requesting the stay establishes, by concrete evidence of the procedural state of the opposition or limitation proceedings, that the decision is likely to be given before, or shortly after, the date of the oral hearing in the proceedings before the Court. In assessing the request the Court shall have regard to the objective that proceedings before the Court be concluded within one year. A mere request for, or grant of, acceleration of the proceedings before the European Patent Office shall not by itself satisfy this paragraph.
295.3 A stay under paragraph 1(d) may be ordered in whole or in part, including a stay limited to one of several patents in suit. Where the parties jointly request a stay, the Court shall order it unless a mandatory provision of these Rules or an overriding requirement of the proper administration of justice requires otherwise.
295.4 The Court shall not order a stay under paragraph 1(m) where the situation is governed by another sub-paragraph of paragraph 1 or by another provision of these Rules. A stay under paragraph 1(m) shall not be ordered against a party with a legitimate interest in obtaining a decision, save where the proper administration of justice clearly so requires.
R.295 lets the Court "may" stay proceedings in eleven situations but supplies almost no decisional standard, producing forum-dependent outcomes on three recurring questions: partial stays, the undefined "rapidly" criterion for EPO-opposition stays, and the residual catch-all in sub-paragraph (m).
The partial-stay question is split on identical (d) wording. Munich (UPC_CFI_815/2024) held R.295(d) "does not provide for a partial stay" and refused to approximate one through case management. Milan (UPC_CFI_472/2024, followed by UPC_CFI_181/2025) held the opposite — a partial stay limited to one patent is possible, and on a joint request "may = shall". New R.295.3 codifies Milan's reading: a joint-request stay may be partial, and is in principle mandatory, while preserving the Mannheim limit (UPC_CFI_210/2023) that the joint-request requirement is strict. Milan's position is preferred because it gives effect to party autonomy and avoids the all-or-nothing trap Munich's reading creates.
New R.295.2 codifies the convergent "rapidly" benchmark: a decision expected before or shortly after the Court's oral hearing (UPC_CoA_237/2025, Juul v NJOY; UPC_CFI_851/2025, Dyson v Dreame), on concrete evidence of the EPO procedural state (UPC_CFI_263/2023, Paris CD), weighed against the one-year-to-hearing target, with mere acceleration insufficient (Dyson v Dreame).
New R.295.4 codifies the Court of Appeal's subsidiarity rule — (m) yields to more specific provisions (UPC_COA_511/2024) — and the Art. 47 Charter "default is not to stay" gloss against a party with a legitimate interest in judgment (UPC_CFI_210/2023).
The proposal numbers the existing list as R.295.1, leaving sub-letters (a)-(m) untouched, so every cross-reference elsewhere in the Rules survives. The three new criteria sit in new paragraphs rather than rewriting the sub-letters, keeping the catalogue of stay-grounds and the operative standards textually distinct.
"Before, or shortly after, the date of the oral hearing" tracks the Court of Appeal's own formulation in Juul v NJOY and does not import a fixed number of months — "shortly after" preserves the Court's margin while ruling out the open-ended reading that allowed years-distant EPO dates to support a stay. The "concrete evidence of the procedural state" clause makes the Paris CD evidentiary test (UPC_CFI_263/2023) a condition precedent, not a mere factor.
In R.295.3, "the Court shall order it unless" codifies Milan's "may = shall" without making the duty absolute: the "mandatory provision / overriding requirement" carve-out preserves the Mannheim qualification. One might object that "shall" removes all discretion — it does not, because the carve-out and R.295.4 retain it where a co-party's legitimate interest or a specific rule intervenes.
Placing subsidiarity in R.295.4 rather than the Preamble makes it an operative, appealable rule, not interpretive colour, directly enacting UPC_COA_511/2024.
R.370.9(b) (amendment); new R.370.9(b-bis) and new R.370.9(f) (insertion); R.360 (amendment); new R.370.2(j) (insertion). Operations: extend the existing withdrawal-reimbursement rule to PI/preliminary applications and to R.360 mootness disposals, add an express fee item and reimbursement track for FRAND/licence counterclaims, and align R.360's wording with R.370.
Rule 370.2 — fixed fees (insertion of new sub-paragraph (j); existing (a)–(i) unchanged):
[R.370.2(a)–(i) unchanged]
(j) Counterclaim for a licence on fair, reasonable and non-discriminatory (FRAND) terms raised in answer to an infringement action [Rule 25, Art. 32 (1) (a) UPCA].
Rule 370.9 — reimbursement (amendment of (b); insertion of (b-bis) and (f); (a), (c), (d), (e) unchanged):
(a) (unused item)
(b) In case of the withdrawal of an action [Rule 265] or of the withdrawal of any application listed in Rule 370.2 (f) to (i) [applications for provisional measures, to preserve evidence, for an order for inspection or to freeze assets] the party liable for the Court fees will be reimbursed by 50% if the action or application is withdrawn before the closure of the written procedure. Where no written procedure is held, the relevant cut-off shall be the closing of the oral hearing on the application or, if none is held, the date of the order disposing of the application.
(b-bis) Where the Court disposes of an action, counterclaim or application by way of order under Rule 360 because it has become devoid of purpose, the party liable for the Court fees will be reimbursed by 50% if the event rendering the matter devoid of purpose occurred before the closure of the written procedure, applying the cut-off in paragraph 9 (b) for matters in which no written procedure is held.
(c) [unchanged]
(d) [unchanged]
(e) [unchanged]
(f) The reimbursements provided for in paragraphs 9 (b) and (b-bis) shall apply to the fee paid under Rule 370.2 (j) for a FRAND-licence counterclaim, and shall apply mutatis mutandis to the withdrawal of an appeal, the relevant cut-off being the closure of the written procedure on appeal.
Rule 360 — No need to adjudicate (amendment):
If the Court finds that an action , counterclaim or application has become devoid of purpose and that there is no longer any need to adjudicate on it, it may at any time, on the application of a party or of its own motion, after giving the parties an opportunity to be heard, dispose of it the action by way of order. A disposal under this Rule shall give rise to reimbursement in accordance with Rule 370.9 (b-bis).
The R.370 fee schedule and its reimbursement rule (R.370.9) are keyed to a closed list of "actions" and speak only of the "withdrawal of an action [Rule 265]" or Centre settlement. Three recurring procedural events fall outside that wording, so divisions fill each by analogy, expressly labelling the omission a "planwidrige Regelungslücke." (1) PI-application withdrawals: Munich applies R.370.9(b) analogously (UPC_CFI_755/2024 — "Es besteht eine Regelungslücke"; UPC_CFI_515/2024 — direct application excluded but analogy appropriate because the gap is planwidrig), as do Nordic-Baltic and The Hague. (2) Counterclaims mooted under R.360: UPC_CFI_300/2023 applies R.370.9(b)/(c) by analogy to a counterclaim mooted by EPO revocation, and UPC_CFI_249/2023 applies R.360 itself by analogy to mooted PI proceedings. (3) FRAND-licence counterclaims carry no fee in the R.370.2 table: UPC_CFI_487/2023 holds the absence "an unplanned gap in the Rules of Procedure." The Court of Appeal endorses the technique: UPC_CoA_611/2024 (DISH v Aylo) and UPC_CoA_641/2025 (Sanofi/Regeneron v Amgen — R.370.9(b)(i) applies "by analogy to appeal withdrawals").
There is a narrow but real split on PI reimbursement: Munich/Nordic-Baltic/The Hague refund by analogy; Milan refuses (UPC_CFI_472/2024 — reimbursement "is limited to actions … and does not apply to applications for provisional measures, either directly or by analogy"). I codify the majority/CoA-endorsed position because the Court of Appeal has already validated the analogy technique (UPC_CoA_611/2024, UPC_CoA_641/2025) and the incentive rationale is identical for actions and applications; the amendment simply makes the settled analogy textual and uniform.
I amend R.370.9(b) rather than create a free-standing rule so the existing 50%/written-procedure architecture and its cut-off carry over unchanged — the divisions already read the rule this way, so editing it (not bolting on a parallel regime) minimises interpretive friction. The explicit cross-reference to R.370.2(f)–(i) names exactly the provisional/evidentiary applications the case law treats as covered, no more, closing the Milan objection that "applications" differ from "actions" by enumerating them. A new (b-bis) is the right home for R.360 mootness because mootness is conceptually distinct from voluntary withdrawal yet should attract the same quantum; I tie the trigger to when the mooting event occurred, not the disposal date, so a party is not penalised for the Court's docket timing. FRAND fees go in a new R.370.2(j) plus a reimbursement hook in (f): the table must first price the step before reimbursement can attach. Counter-reading: one might object that adding R.370.2(j) imposes a new fee burden on SME defendants. It does not set the amount — that stays with the Administrative Committee's table of fees under Art. 36(3) UPCA and the R.370.8 SME 50% discount applies automatically; the rule only makes a priced, predictable step out of one that divisions currently improvise. I append the appeal-withdrawal analogy (UPC_CoA_641/2025) to (f) rather than R.370.5 to keep all reimbursement logic in one paragraph.
R.151 introductory paragraph (amendment — deadline trigger on appeal); new R.151(f) (insertion — appeal-superseded recovery flag); new R.152.4 (insertion — interest on awarded costs); new R.152.5 (insertion — recoverability of cost-proceeding costs).
R.151 — Start of proceedings for cost decision
Where the successful party (hereinafter "the applicant") wishes to seek a cost decision, it shall within one month of service of the decision — or, where a first-instance cost assessment is set aside or varied on appeal, within one month of service of the decision of the Court of Appeal — lodge an Application for a cost decision which shall contain:
[R.151(a)–(c) unchanged]
(d) an indication of the costs for which compensation is requested, which may include recovery of court fees and costs of representation, of witnesses, of experts, and other expenses; the costs of the cost proceeding itself shall be recoverable only to the extent provided in Rule 152.5; and
(e) the preliminary estimate of the legal costs that the party submitted pursuant to Rule 118.5; and
(f) where the Application is lodged following a decision of the Court of Appeal that has set aside or varied a first-instance cost assessment, a statement to that effect and an indication of any amount already paid under the assessment so set aside or varied, recovery of which may be sought in the proceedings.
R.152 — Compensation for representation costs
[R.152.1–.3 unchanged]
4. Interest shall not accrue on costs awarded under Rule 151 by operation of these Rules. A party who seeks to secure recovery against delay may apply for provisional recovery of costs in accordance with Rule 221.1(d) and Rule 242.2(a).
5. A party who is successful in the proceedings for a cost decision shall bear its own costs of those proceedings, save that the court fee of any appeal in the cost proceedings shall be recoverable.
R.151 lists what a cost application "shall contain" and fixes a one-month deadline, but is silent on three recurring questions: whether interest accrues on the sum awarded, whether the costs of conducting the cost proceeding are themselves recoverable, and which "decision" starts the month where a first-instance assessment is later superseded on appeal. The case law has settled all three, so the Rule should make the settled position visible.
Interest: Munich holds the omission deliberate — because interest is expressly provided for damages (R.131.2(a)) but not in R.151, there is "no unplanned legislative gap" (UPC_CFI_249/2023; UPC_CFI_696/2024), the remedy for delay being provisional recovery under R.221.1(d)/R.242.2(a). R.152.4 codifies exactly that line, including the signpost to the provisional-recovery remedy.
No costs upon costs: the Court of Appeal holds a successful party in R.150 cost proceedings bears its own cost-proceeding costs, only the court fee being recoverable (UPC_CoA_618/2024, Hanshow v VusionGroup; UPC_CoA_153/2025, Tiroler Rohre), followed by the LDs (UPC_CFI_335/2025, Mannheim; UPC_CFI_355/2023, Düsseldorf). The carve-out's boundary is articulated differently — Mannheim limits it to "the court fee of an appeal" while Hanshow speaks of "the court fee." R.152.5 adopts Mannheim's narrower, more precise formulation, since the broader reading would let a party recover the first-instance cost-proceeding fee the CoA never endorsed.
Deadline: the CoA runs the month from service of the CoA decision and requires a separate application for appeal costs (UPC_CoA_1/2024), and Düsseldorf holds a CoA reversal strips the first-instance assessment of its basis, allowing recovery of already-paid sums in fresh post-appeal proceedings (UPC_CFI_1110/2025, Seoul Viosys). The amended chapeau and new R.151(f) codify both.
Interest sits in R.152, not R.151, because R.151 governs the application's contents whereas R.152 already governs the quantum of recovery — interest is a quantum question, so it belongs alongside the recoverability rules. The negative phrasing ("Interest shall not accrue … by operation of these Rules") tracks the deliberate-silence holding precisely: it does not bar interest a party secures by other means, only confirms the Rules create none. The signpost to R.221.1(d)/R.242.2(a) is included because the courts identified provisional recovery as the substitute remedy; omitting it would leave the prejudiced party without the cure the case law supplies.
R.152.5 codifies "no costs upon costs" with the narrow appeal-fee carve-out. One might object that "the court fee of any appeal" is narrower than Hanshow's "the court fee"; that is deliberate — the CoA's own holding concerned an appeal fee, and the wider reading rests on dicta, so the operative text should not extend further than the ratio.
The deadline change is split between the chapeau (the alternative trigger, a true amendment to the time-limit) and new R.151(f) (a contents requirement flagging the superseded assessment), keeping each provision functionally coherent. R.151(e) gains "; and" because (f) now closes the list; no existing sub-letter is renumbered.
R.323.1, R.323.2 (amendment); new R.323.2a (insertion of an applicant reply); R.19.1(c) (amendment — clarifying note).
R.19.1(c) (amended):
[R.19.1(a) unchanged]
[R.19.1(b) unchanged]
(c) the language of the Statement of claim [Rule 14], limited to compliance with Article 49(1) and (2) of the Agreement. An application to use the language in which the patent was granted on grounds of fairness pursuant to Article 49(5) of the Agreement is not a Preliminary objection and shall be made in accordance with Rule 323.
R.323.1 (amended):
R.323.2 (amended):
R.323.2a (new — insertion):
2a. The applicant may lodge a reply to the position indicated under paragraph 2, limited to matters raised in that position, within 7 days of service of the position. No further written submission shall be made unless the President, having regard to the need to decide the language of the proceedings expeditiously, grants leave.
[R.323.3 unchanged]
A defendant (or claimant) who wants the proceedings conducted in the language the patent was granted in has two ill-fitting routes. The R.19.1(c) preliminary objection cross-refers only to Rule 14, so courts read it as reaching only compliance with Art.49(1)–(2) UPCA, not the broader Art.49(5) fairness/convenience question; that question must instead go through R.323, which names only the Application and a single 10-day respondent position — and no applicant reply.
The case law on both points is settled and consistent (no division split). On R.19(c)'s narrow scope: UPC_CFI_583/2024 (Paris LD) holds an Art.49(5) fairness challenge is not a R.19(c) objection. The exhaustiveness of the R.19.1 grounds list is confirmed up and down the Court: CoA UPC_CoA_188/2024, Munich LD UPC_CFI_15/2023, and Munich CD UPC-CFI-2296/2025. On R.323's missing reply: the CoA held squarely in UPC_CoA_354/2024 (Apple) that "in R.323 RoP there is no reply to a respondent's statement," and the Brussels CFI President followed in UPC_CFI_871/2026 (Esko v In(k)control), treating unsolicited further submissions as not to be considered. On timing, courts had to invent the rule themselves: UPC_CoA_207/2024 and UPC_CFI_239/2023 (CFI President) read R.323.1's "shall include in the Statement of Defence" as a deadline, not a fixed filing moment. I codify the CoA-led consensus: I make the timing an explicit deadline (R.323.1), and — because the present no-reply rule leaves the single 10-day position decisive and pressures applicants to front-load everything — I grant a short, scoped reply (R.323.2a) rather than leaving it to ad hoc leave.
I add the reply as a new R.323.2a rather than amending R.323.2, so the respondent's 10-day position and the applicant's reply remain distinct, separately-clocked steps and existing R.323.3 need not renumber. The reply is "limited to matters raised in that position" and capped at 7 days to preserve the expedition the CoA prized in UPC_CoA_354/2024 — it answers the no-reply gap without converting a language incident into full-blown pleading exchange. The residual leave clause codifies the very standard the courts already apply (UPC_CFI_871/2026), so nothing further is filed as of right.
Counter-reading: one might object that granting a reply reopens the door the CoA deliberately closed in Apple. It does not — Apple decided that the current text affords no reply, not that fairness forbids one; the President's leave-gated discretion over anything beyond the reply is preserved verbatim. A second objection: why touch R.19.1(c) at all if courts already read it narrowly? Because the narrowing is judge-made and invisible on the face of the rule. Adding "limited to compliance with Article 49(1) and (2)" plus a signpost to R.323 makes the boundary operative and self-explanatory, sparing future defendants the wasted, out-of-time R.19(c) filings the case law shows them attempting. The R.323.1 timing words mirror UPC_CoA_207/2024's "outer limit" reading without inventing new doctrine.
R.220.3 (amendment — insert substantive standard); new R.220.3a (insertion — cumulative/alternative structure of the gateway); R.333.1 (amendment — clarify panel competence over judge-rapporteur access orders under R.262.1(b)).
Rule 220.3
Rule 220.3a (new)
3a. The grounds in Rule 220.3(a) and (b) are alternative. A request founded on Rule 220.3(b) alone may be granted, notwithstanding the absence of manifest error, where the requesting party demonstrates a genuine divergence in the practice of the divisions of the Court of First Instance on the question of law concerned.
Rule 333.1
[R.333.2 – .5 unchanged]
R.220.3 names the discretionary-review remedy and its 15-day deadline but states no substantive criterion for granting leave; R.333.1 entitles a party to panel review of "case management decisions or orders" without defining whether a judge-rapporteur's access decisions fall within it. The Court of Appeal has filled the first gap with a judge-made two-pronged gateway — the order must be (i) manifestly erroneous, or (ii) raise a fundamental question of law whose review is necessary for the consistent application of the Rules — drawn from Preamble point 8, which is not part of any operative rule (UPC-CoA-74/2026, ASC v Motorola; UPC-COA-0000056/2026). Mere alleged incorrectness fails and arguments not put below are barred (UPC_CoA_489/2024; UPC_CoA_570/2024). But standing judges split twice. On structure: UPC-CoA-67/2026 reads the prongs as cumulative ("requires both"), while UPC-COA-0000056/2026 and UPC-CoA-35/2026 read them as alternatives. On the consistency prong standing alone: UPC-COA-0000052/2026 grants review on divergence alone, whereas UPC_CoA_469/2024 demands case-specific misuse of discretion in addition. On R.333 scope, Munich (UPC_CFI_1234/2025) denies panel competence over a judge-rapporteur's R.262.1(b) access decision, while Mannheim (UPC_CFI_471/2023) treats R.262A/access orders as squarely reviewable. I codify the alternative reading (the majority CoA line, UPC-COA-0000056/2026 over the isolated UPC-CoA-67/2026), the divergence-suffices position for the consistency prong (UPC-COA-0000052/2026), and Mannheim's broad R.333 scope — restoring predictability to which procedural track a party must use.
The standard is placed in the body of R.220.3 rather than left in Preamble point 8 because the Preamble is interpretive, not operative or directly reviewable; a litigant must be able to read the gateway off the rule that governs the request. The drafting tracks the CoA's own verbatim formulations — "manifestly erroneous", "fundamental question of law", "consistent application and interpretation" — so it codifies, not invents. The cumulative-versus-alternative split is the live ambiguity, so it is resolved explicitly in a dedicated R.220.3a rather than buried in 3: the word "alternative" and the "notwithstanding the absence of manifest error" clause leave no room for the UPC-CoA-67/2026 cumulative reading. A reader might object that the divergence requirement reopens a second instance for any contested point of law; it does not, because R.220.3(b) still requires a "fundamental question" and 3a still requires a "genuine divergence in the practice of the divisions", excluding ordinary disagreement (UPC_CoA_570/2024). The R.333.1 insertion names R.262.1(b) and R.262A orders specifically, the precise category Munich and Mannheim split on, rather than restating "broadly" — naming the disputed orders is what makes the boundary operative. No existing paragraphs are renumbered; "3a" is inserted between 3 and 4 to keep the access-route and the gateway-standard adjacent.