COMVIK: examining an invention that combines technology and a business rule

A phone, a SIM and a charging arrangement

COMVIK's patent concerned a GSM system in which a user selected an identity allocated to a SIM. The claims combined infrastructure operations with allocating business and private call costs or costs between users. The patent was revoked after opposition. On 26 September 2002 the Board dismissed the appeal for lack of inventive step (facts I–V; reasons 13–17). ^[1]

The combination remains instructive. A useful business arrangement may operate through complex infrastructure. Implementation difficulty and commercial usefulness do not yet locate a technical contribution. Analysis requires an identified feature, an effect and a comparison with earlier solutions.

Two levels of a mixed claim

COMVIK permits technical and non-technical features in one claim. Inventive-step assessment considers features contributing to technical character; others cannot by themselves establish inventive step. A non-technical aim may instead be a given constraint within the technical problem (headnotes I–II; reasons 4–8). ^[1]

Articles 52 and 56 EPC provide the legal framework. Current EPO Guidelines also describe features which appear non-technical in isolation but contribute, in context, to a technical effect serving a technical purpose. Therefore do not mechanically delete every element called an algorithm, or assume that using a computer always establishes inventive step. ^{[2] [3]}

Editorial method: from business requirements to a technical effect

Our proposed discussion with a team begins with two descriptions. The first states the organisational outcome: who pays, what is selected and which rule determines allocation. The second describes implementation: data structures, messages, device states, timing and resources. Separating the descriptions does not mean the features are independent; it makes their interaction examinable.

Then create a matrix of claim features. For each, identify the supporting disclosure, asserted effect, verification method and possible earlier document. Instead of saying faster system, specify what becomes faster, under which conditions and against which baseline. Where measurements are missing, distinguish technical reasoning from a hypothesis requiring further work.

Next examine novelty and non-obviousness separately. As an editorial exercise, formulate the problem without inserting the essential solution. Asking how to use our particular memory structure already includes part of the answer; asking about a defined access conflict permits an investigation. Treating a business requirement as a constraint does not authorise smuggling the inventive technical feature into the problem.

Example: reservations with limited computing resources

A hypothetical platform prioritises a category of customers. The priority rule describes an organisational choice. Its team also claims that a new synchronisation mechanism reduces conflicting writes between distributed nodes. That assertion needs separate architectural records, concurrent-event analysis and failure scenarios. We do not assume that the mechanism is actually new.

A comparison can use the same requests, memory limits and network conditions for the reference and experimental implementations. The report should disclose costs as well, such as more messages or different consistency conditions. The exercise does not replace a prior-art search or analysis across the full claim scope. It distinguishes a benefit to the business model from the proposed behaviour of a technical system.

From SIM cards to AI: what this case can contribute

Noam Shemtov discusses AI protection in Beyond algorithms: patenting core AI innovation in Europe, in A Research Agenda for Patent Law, 2025, 65–80, DOI 10.4337/9781803927855.00012. The public abstract addresses tensions between AI innovation, technical character and COMVIK. We cite the bibliography and accessible abstract, without treating the chapter as an up-to-date account of every subsequent judgment. ^[4]

The editorial lesson for AI and software projects concerns collaboration. Before drafting, researchers, engineers and the person assessing patentability should agree on the language describing the solution. Intelligent, automatic and optimal need translation into features and testable relationships. COMVIK supplies no predetermined result for a new product; it structures questions requiring technical evidence and prior-art analysis.

Record the results both as a short decision map for the team and a detailed evidential annex. The shorter account should retain conditions on which the technical conclusion depends. If an effect has been shown only at a certain load or configuration, keep that limit when discussing the proposed scope of protection. Otherwise an executive summary may promise more than the evidence supports.

Sources and further reading

1. 1
   EPO, Two identities/COMVIK ↗

   T 641/00; ECLI:EP:BA:2002:T064100.20020926; 26.09.2002; OJ EPO 2003, 352

   Primary decision: headnotes I–II, reasons 4–8 and 13–17. Official metadata: epo.org/en/boards-of-appeal/decisions/t000641ex1.

2. 2
   European Patent Convention ↗

   EPC, art. 52 and 56

   Articles 52 and 56 EPC; Article 56 on inventive step is available in the same official edition.

3. 3
   EPO Guidelines — Claims comprising technical and non-technical features ↗

   Guidelines for Examination 2026, G-VII, 5.4

   Official explanation of COMVIK, assessment of features in context and the problem-solution approach; checked 18 October 2026.

4. 4
   Noam Shemtov, Beyond algorithms: patenting core AI innovation in Europe ↗

   A Research Agenda for Patent Law, Edward Elgar, 2025, 65–80; DOI:10.4337/9781803927855.00012; eISBN:9781803927855

   Publisher metadata and public abstract verified. Academic context for AI, not a substitute for the decision or later legal updates.