IP-Fälle und Artikel

Patenting for the Cloud - Challenges and Solutions

In this article we look at the challenges in securing IP that arises from the distributed nature of cloud computing technology. Specifically, we look at the situation where there is clear potential to fragment a cloud-based invention into stages which can be shared between servers that are in different countries and/or run by different subsidiaries and affiliates. All of these factors can frustrate the process of obtaining enforceable patent protection for the invention as a whole. In particular, such fragmentation means that any particular third party might only implement a subset of the steps in a patented algorithm in any one jurisdiction.

Legal solutions

If a third party does not infringe all the claimed steps of an algorithm (or the features of any cloud-based process or system) then it may still be possible to bring an action against them on the basis of so-called contributory or 'indirect' infringement provisions.

On this basis it may also be possible in the UK to bring together several partially infringing parties under the principle of 'joint tortfeasorship' (ie, collusion in a civil wrongdoing). This is of particular interest given the disproportionately large number of Europe's cloud computing facilities found in the UK (see note 1 below). This principle also appears to have been effectively endorsed by the US Federal Circuit in Akamai v Limelight, where it has recently ruled that "it is no longer necessary to prove that all the steps were committed by a single entity".

However, contributory infringement provisions are not a panacea. They are typically weaker rights than those for primary or 'direct' infringements and, depending on the jurisdiction, may come with strings attached:

  1. they can impose requirements to demonstrate wilful infringement;
  2. they can require agreement on the extent to which the local infringement is essential to the invention as a whole; and
  3. they may not apply to the production of exports from the jurisdiction (eg, the processed data) – which is likely to be relevant in a cloud-based implementation of an algorithm.

These factors add a layer of uncertainty to a patent's value that it is clearly preferable to avoid.

Meanwhile, a new and potentially significant legal solution in Europe arises from the agreement on 11 December 2012 to a unitary European patent (see note 2 below) covering 25 European states (see note 3 below), which if ratified will become another option alongside existing national and conventional European patents.

The Unitary Patent might help to address the distributed nature of cloud based inventions by treating the majority of Europe as a single jurisdiction for the purposes of infringement. When coupled with Europe's strict data protection laws, which make the export of user data outside of Europe relatively difficult, this makes the direct enforcement of many cloud-based patents within Europe highly feasible.

However, the Unitary Patent does not address every issue:

  1. It clearly only applies within Europe, and so any cloud network extending beyond the participating states may again only partially infringe a unified patent.
  2. It of course does not help in other parts of the world where there are well established multi-jurisdictional networks, such as the US/Canada.
  3. It also does not clearly address the fact that, even within a single jurisdiction, different steps of an algorithm may be implemented by different legal entities, so that the weaker contributory infringement provisions need to be relied upon.

The ideal solution would therefore seem to be to file multiple applications, each directed to separate subsections of the algorithm that are likely to be implemented together, so that a realistic implementation of the algorithm is likely to wholly infringe at least one, and preferably several, such patents.

The problem with this mosaic approach is that (in addition to being expensive) the more one breaks an inventive concept down, the less inventive the constituent parts tend to be. This risks creating a collection of patent applications that are all unenforceable in practice. In other words, it creates a trade-off between inventiveness and the ability to pursue direct rather than indirect infringement.

Drafting solutions

The good news is that this is not a new problem. The mobile phone industry for example relies upon a co-operative relationship between phones, base stations, and networks, in which each may be owned by a different party or located in different countries, and in which a new invention is likely to affect devices of multiple parties to a greater or lesser extent in order to work. Coupled with the fact that all too often the most economically valuable element in the system may not, in isolation, contain the most inventive subset of features, this can again result in the need to trade-off between inventiveness and infringement in a similar manner.

In response, an independent claim to an isolated element of a system (or a subsection of a cloud algorithm) can be drafted to pull in as much context from the rest of the invention as possible in a non-limiting manner, to tip the trade-off in the patentee's favour.

For example, if the focus of an invention is an improvement to a base-station transmitter, with only a small modification needed to the mobile receiver, then a claim to that receiver may relate to how it is adapted to take advantage of (or enable) the specific improvements in the transmitter.

In the cloud, the use of the past tense and passive voice can similarly help; a claim to a processing step that operates with data that has had a special process already applied to it incorporates the special character of the data into the claim, but avoids incorporating the step that applied the special process. This is useful if that step was performed on a cloud server in another country. Other drafting methods include including the broader system as something that the claimed step is suitable for operating with, or the preferential use of method claims that can be more easily written to be independent of any particular device.

In these ways it is possible to boost the inventiveness of claims in a fragmented cloud algorithm, without unduly limiting them. The result is hopefully a set of patents that are directly enforceable locally whilst being robust.


A final issue with cloud applications is that, unlike in telecoms, different stages of an invention are typically less restricted to specific hardware. As a result the number of different possible implementations of an algorithm over multiple cloud servers can become very large.

For an example algorithm that could be separated into three sections (A, B C), potentially twenty seven implementations of the algorithm may need to be protected. In figure 1a, 'R' means a device (ie, a server) is not implementing a step of the claimed algorithm, but might be in communication with a device that is.

In fact, in this idealised case one could write an independent claim to a server performing at least one of sections A, B and C and handling, as appropriate, data for those sections of A, B and C it does not perform but which it receives from or transmits to a further server. Arguably this could cover any of the three servers for all combinations of A, B, and C.

However, the situation gets more complex when only some permutations are possible. In a notional example section A relates to a 'front end' stage of the algorithm that can only be implemented by servers 1 and/or 2, whilst sections B and C are always implemented separately. An example of such a set-up may be a voice compression, recognition and archive/audit system for a phone banking service.

Fortunately, the more complex configuration of possible permutations of the algorithm can be treated under Rule 43(2)(c) of the European Patent Convention as a set of alternate solutions to the same problem, allowing in this case three independent claims to cover the cloud algorithm within a single patent application, along the lines of:

  1. device 1 or 2 performing step A and a respective one of steps B and C;
  2. any device performing one of steps B and C, depending on whether data for complementary steps AC or AB are available/required as applicable; and
  3. any one of the devices for a system where each device performs a respective one of steps A, B and C, excluding that device 3 performs step A.

Hence by using multiple independent claims to logically segregate configurations of the invention, it is still possible to cover a complex pattern of implementations methodically and accurately.

In summary, whilst cloud-based inventions do present challenges in a patent system that never envisaged the protection of internationally distributed processes, the combination of existing direct and indirect infringement provisions and new unitary patent protection, together with carefully balancing inventiveness with local enforcement, means that innovations in the cloud should be able to enjoy patent protection like any other facet of technology.

Notes and Useful Links:

  1. See http://dycip.com/datacentremap
  2. See http://dycip.com/epo-unitarypatent
  3. Currently excluding Spain and Italy