Is 'Software' Too Vague to Protect? EUIPO Refers Landmark Question to the Grand Board
Background: A Dispute Between AHEAD Automotive and Deutsche Bahn
Deutsche Bahn AG applied to register the figurative sign KIRA for, among other things, autonomous-driving software in Class 9 and vehicles, including driverless vehicles in Class 12. AHEAD Automotive GmbH opposed on the basis of its earlier EU mark QIRA, registered for computer software and related services in Classes 9, 35, 41 and 42 (but notably not for any goods in Class 12).
The Opposition Division rejected the opposition in relation to the Class 12 goods, holding that vehicles and software-related goods and services differ "in nearly every respect": different purpose, different use, different distribution channels, different producers, and no complementary or competitive relationship. The mere fact that software is integral to many products, including cars, was not enough; otherwise, the Division reasoned, virtually everything would be similar to software.
AHEAD Automotive appealed, pointing to a line of Board of Appeal decisions finding software similar to vehicles (Vdl e-power/e-POWER, R-2914/2019-1; Xs direct/Xs, R-2387/2020-2; Xs/Xs, R-464/2021-2). Deutsche Bahn countered with contrary authority (PONS/PONS SEGURIDAD VIAL, R 1160/2022-4; the Opposition Division's Smart D (fig.)/smart (fig) decision), arguing that vehicle manufacturers develop software in-house for their own products rather than marketing it independently, and that the relevant publics simply don't overlap.
Two Questions of Fundamental Importance
The Board found genuinely conflicting Board of Appeal case law on goods similarity, which is the fundamental requirement for a referral to the Grand Board under Article 165(3) EUTMR and Article 37(1) EUTMDR. However, the Board of Appeal went even a step further and identified a logically prior issue:
1. Is "software" (and "SaaS") sufficiently clear and precise under Article 33(2) EUTMR?
Under current EUIPO practice, "software" is an accepted term drawn from the Harmonised Database and is not objected to. It appears in neither the Common Communication CC1 (on unacceptable classification terms) nor CC15 (on lack of clarity/precision generally). Yet Advocate General Szpunar, in his Opinion in Sky (C-371/18), took the opposite view, suggesting that a mark registered for "computer software" as such confers a monopoly of "unjustified" scope and questioning why the EUIPO's own concerns about the breadth of "machines" in Class 7 shouldn't apply equally to software. The CJ's judgment in that case did not discuss this point.
2. If the term is clear, are software and vehicles similar goods?
Only once (and if) the first question is resolved does the underlying, and already diverging, similarity question become relevant.
The Bigger Picture: A New IP Translator?
This is where the referral becomes interesting well beyond the parties. Software is no longer a discrete product category. Obviously, it is embedded in almost every electronic and digital device on the market, and it underpins a large share of the services rendered for example in the financial, business, IT and telecoms sectors. If the Grand Board (and, after that, the General Court and maybe even the Court of Justice [*wishful thinking*]) were to follow AG Szpunar's line and hold that a bare "software" specification lacks the clarity and precision Article 33(2) EUTMR demands, the fallout would extend far beyond one autonomous-driving trade mark dispute. It would put a question mark over a very large number of existing Class 9 and Class 42 registrations - a scenario with more than a passing resemblance to IP Translator (C-307/10), which forced an EU-wide reassessment of how class headings are to be interpreted and in fact even a change in law and transitional rules to accommodate trade marks owner who had registered their marks under the old regime.
There is a case, though, for seeing this as a welcome correction rather than a disruption. Class 9 is notoriously overcrowded, not least because "software" has become a default, catch-all term. And conceptually, the critique has force: software is, by definition, a set of instructions enabling a computer to perform a task, but the specific purpose of one type of software (say, autonomous-vehicle control systems) is nothing like that of another (say, HR analytics tools). Treating all "software" as a single, undifferentiated category for similarity purposes arguably obscures more than it clarifies, and a more granular approach (consistent with the EUIPO's own Guidelines on Information Technology goods and services, which already distinguish generic software from software specifically designed to operate a particular apparatus) could bring welcome discipline to both classification and opposition practice.
What This Means in Practice
- Watch this referral closely. Its outcome will resolve not just the Class 9 vs. Class 12 similarity question but potentially reopen how "software" specifications are drafted and examined across the board.
- Review broad software specifications now. Rights holders relying on bare "computer software" or "SaaS" terms in Class 9 and 42 should consider whether more granular, purpose-specific wording would better withstand a stricter clarity standard and better support similarity arguments in future disputes.
- Pending oppositions and appeals involving software vs. specific goods (vehicles, machinery, medical devices, etc.) may be worth staying, or at least flagging the Grand Board referral, given the currently unsettled case law the Board itself identified.
- New applications covering both software and the apparatus it controls should consider explicit purpose-based drafting to future-proof the specification against either outcome.
We Can Help!
Whether you are drafting a new Class 9 / 42 specification, defending an existing software portfolio, or assessing an opposition that turns on software-versus-goods similarity, the coming Grand Board decision could materially affect your position. Our trade mark team is happy to review your filings in light of this referral and help you plan ahead.
You might also be interested in this
In 2019, in an article for the Festschrift honoring Paul Ströbele, I examined how companies can defend themselves when their trademarks are used without consent as eye-catching vehicles for political messages.
OLG Bamberg, judgment of 29 July 2026 – 3 UKl 13/25 e | The Third Civil Senate of the OLG Bamberg continues to develop its Digital Services Act case law. For the first time, it specifies how far providers must disclose the weighting of their recommendation system parameters; at the same time, it prohibits login requirements as a precondition for reporting illegal content. We summarise the decision and provide guidance for practice.
The General Court also dismisses OpenAI’s action: No EU trademark protection for the word mark “OPENAI”—at least for now.
LG München I, judgment of 31.07.2026 – 42 O 763/25. In its second landmark judgment on the relationship between copyright and AI training, the Munich Regional Court (LG München I) has found the operator of the AI music generator Suno liable for infringement of copyright in six popular musical works. Notably, the 42nd Civil Chamber has, for the first time in a German judgment, applied the US fair use doctrine to AI training – and rejected it. We summarise the decision and set it in context.








