Novelty over Generic Disclosures

For a claimed invention to be novel, it must have at least one feature that is not disclosed in a single piece of prior art, typically a document. As described in the Guidelines for Examination, G-VI 2:

"A document takes away the novelty of any claimed subject-matter derivable directly and unambiguously from that document including any features implicit to a person skilled in the art in what is expressly mentioned in the document, e.g. a disclosure of the use of rubber in circumstances where clearly its elastic properties are used even if this is not explicitly stated takes away the novelty of the use of an elastic material. The limitation to subject-matter "derivable directly and unambiguously" from the document is important. Thus, when considering novelty, it is not correct to interpret the teaching of a document as embracing well-known equivalents which are not disclosed in the documents; this is a matter of obviousness".

Novelty is therefore not quite as straightforward as simply looking through the document and deciding whether all features in the claim can be found. But what are implicit features? The above section of the Guidelines makes clear that this does not extend to well-known equivalents, which would instead fall under inventive step. There is, however, a grey area between an explicit disclosure of something and what would constitute an equivalent. 

A question I have been asking myself recently is whether implicit features would include something as simple as left and right handedness if a disclosure does not specify either but from which the skilled person would inevitably select one or the other, perhaps with equal probability but perhaps not (not a mixture, anticipating clever comments from chemists about enantiomers). If, to take as simple an example as I can think of, a document discloses a widget with a screw but does not disclose whether the screw is a left-handed or right-handed screw (and the handedness doesn't make any difference), does this anticipate a later claim to the exact same widget with a right-handed screw? 

Another section of the Guidelines, which I find has remained unchanged for at least the past 20 years, is G-VI, 5, which states: 

"In considering novelty, it is to be borne in mind that a generic disclosure does not usually take away the novelty of any specific example falling within the terms of that disclosure, but that a specific disclosure does take away the novelty of a generic claim embracing that disclosure, e.g. a disclosure of copper takes away the novelty of metal as a generic concept, but not the novelty of any metal other than copper, and one of rivets takes away the novelty of fastening means as a generic concept, but not the novelty of any fastening other than rivets".

This section of the Guidelines goes back at least as far as the 2001 edition (see the EPO archive versions), where the same wording was used in section C-IV, 7.4, and most probably much earlier than this. A puzzle I am having with this, however, is that this does not refer to any case law. I am also unsure what it means by using the word "usually". Looking in the EPO's case law book, section 5.2.6 on generic disclosure states:

"It is established case law that a specific disclosure destroys the novelty of a generic feature in a claim but that a generic disclosure does not destroy the novelty of a specific feature (see T 651/91, T 776/07, T 1174/05, T 6/04, T 776/07). A generic disclosure does not normally deprive a claimed specific embodiment of novelty (T 1786/09).

In T 651/91 the board pointed out that a generic disclosure did not normally deprive any specific example falling within that disclosure of novelty. A disclosure could be generic even where it only left open the choice between two alternatives".

The key decision therefore appears to be the earliest one mentioned here, T 651/91, which was decided in 1993. The case law book appears to make it clear that a disclosure could be generic with only two alternatives, meaning that my example above of an undefined handedness screw would not anticipate a screw with a specified handedness. Or would it? The Board only states that a disclosure could be generic if there were two alternatives, and the Guidelines states that a generic disclosure does not usually take away the novelty of a specific example. What does this decision actually say though, and what is the legal reasoning behind it? The relevant part of the decision is in points 4.3 and 4.4, which (in machine translation from the original German) state:

"4.3. Consistent with the Guidelines for Examination at C-IV, 7.4, the disclosure of a general teaching does not pre-empt the novelty of a particular practice falling under that general teaching. A teaching can be general even if it leaves open the choice between only two alternatives.

4.4. In the present case, however, more than two alternatives are included. Contrary to the opinion of the appellant, according to which the word sequence "one of the ... gears" in claim 1 of publication (D1) represents a reference to only two gear designs, one of which corresponds to Figure 1 there and the other to the gear type according to the contested patent corresponds, no such restriction can be found in the overall documentation of publication D1. As US Pat. No. 4,280,583 (D3) shows, the general teaching according to claim 1 of document D1 also includes at least the further embodiment shown there, in which three gear wheels are seated on the countershaft, one of which controls the drive member ( 8, 12) of the differential gear drives" (emphasis added).

Note that, as mentioned above, the section of the Guidelines referred to has been unchanged since at least 2001, and most likely also since the date of this decision (it would be useful to have confirmation of this). The result is that we appear to have legal reasoning from the leading case on the subject, which is referred to approvingly in 18 later cases, but which has only a reference to the Guidelines in support of its position. From everything I have been led to believe about how these things work, that seems to me to be the wrong way round. Usually the Guidelines will follow the existing case law, not the case law following the Guidelines. 

Furthermore, this leading case explicitly states that the facts of the case do not even relate to there being two alternatives, so the Board cannot have considered the possibility of left and right handedness or other equivalent situations. As far as I can work out (and I have traced the line of reasoning from all cases that refer to T 651/91), there is no case that considers the situation of exactly and only two alternatives, either of which the skilled person would inevitably select. All have followed T 651/91, finding that generic disclosures do not anticipate a specific claim. 

What if the case law is (at least partly) wrong on this point? It would of course need a fact situation to establish this, but one way of supporting an argument that a generic disclosure of a screw would anticipate both left and right handed screws can be found from case law on the subject of implicit features. An example case is T 823/96, which states at point 4.5:

"[T]he Board observes that the term "implicit disclosure" should not be construed to mean matter that does not belong to the content of the technical information provided by a document but may be rendered obvious on the basis of that content. In the Board's judgment, the term "implicit disclosure" relates solely to matter which is not explicitly mentioned, but is a clear and unambiguous consequence of what is explicitly mentioned. Therefore, whilst common general knowledge must be taken into account in deciding what is clearly and unambiguously implied by the explicit disclosure of a document, the question of what may be rendered obvious by that disclosure in the light of common general knowledge is not relevant to the assessment of what is implied by the disclosure of that document" (emphasis added).

Following this, what would be the "clear and unambiguous consequence" of following a disclosure of a widget with a screw? The skilled person would have to select a screw to make the widget, and the screw would inevitably have a handedness. Does it matter which handedness of screw the skilled person selects? Given that right handed screws are far more common, would it only apply to these? What if both were equally possible? At the moment I do not have any answers to these questions, but would be interested to see if there are flaws in the argument that a generic disclosure such as this would actually anticipate both possibilities, notwithstanding the established case law on the subject. 

By the way, I do actually have a fact situation that matches the above and would very much like to see it tested on appeal, but it is not my decision. 

EP3449450 (nChain Holdings) - A lesson in added matter

In a change to my usual practice, this post is aimed more at people who do not necessarily have a detailed knowledge of how the patent system works in Europe, specifically at the European Patent Office (EPO). It might, however, be of interest to those that do because it provides a useful lesson in what not to do when amending an application. First though, some background on why I am interested in this particular patent.

In February 2021 I started looking in detail at patents that were being granted by the EPO at an increasing rate to nChain Holdings Limited, and wrote a post about it here. For those not familiar with the company, it is a part of Craig "Faketoshi" Wright's attempts to claim that he is Satoshi Nakamoto, the pseudonymous creator of Bitcoin. As Chief Scientist at nChain, Mr Wright plays a key role in their patenting activity, and is listed as inventor on most of the applications they have filed over the past 6 years or so. I am not privy to nChain's business model, but it appears to involve getting as many patents as possible and presumably at some point licensing them to others for lots of money. As I mentioned on the Dr Bitcoin podcast earlier this month, their business model actually seems to me more like cargo cult patenting, but I admit that there may be some useful  and valuable inventions somewhere in among the 364 European applications filed so far, it's just that I haven't found any yet. Perhaps tracking cattle will be the killer application that will earn nChain billions, but who knows.

I have been keeping a tally of European patents nChain have been getting granted. At the time of writing it is up to 57, which is quite a lot for a small company, and there are many more in the pipeline. Some may be of passing interest to those working in the general field of blockchain applications, but many appear to be obscure variations on themes already well established in the prior art, if not actually lacking in novelty. Mr Wright has nevertheless made vague threats about various people infringing his patents, for example claiming coverage of the Lightning protocol or NFTs on Ethereum. So far, none of the patents have to my knowledge been raised in infringement proceedings. Surprisingly (to me at least), none of them have been challenged on validity grounds, at least not until I had a go at one of them last year by requesting a UK patent office opinion on one of them, unfortunately without success. 

The general theme of the patents, which turns out to make them quite difficult to attack, is that they tend to claim a collection of features that together is probably novel and include a distinction over the prior art that is arbitrary but plausible to argue on inventive step grounds. If all you were interested in was to get as many patents as possible, nChain's strategy so far would appear to fit the bill perfectly (hence my suspicion of cargo cult patenting). This is actually good news for anyone who might be worried about being sued for infringement, because they would either be very unlikely to infringe any of the patents or, if they were possibly infringing, would be able to easily work around any potentially threatening patent. To put it simply, one patent that nobody can avoid infringing is going to be a lot more valuable than a hundred patents that nobody infringes.

Given the limited resources available at the moment (i.e. my spare time and other people's generosity), a full-scale attack on nChain's patents is impractical and, given the above, arguably pointless anyway. This does not, however, mean that it is not possible to attack them at all. With so many patent applications being prosecuted at once, someone is bound to make a mistake at some point, whether this is the EPO examiner or nChain's attorneys. It just so happens that both of these happened in the case of EP3449450. This application is one of a handful that relate to Turing completeness, which Mr Wright has had a thing about since 2015, claiming that the Bitcoin protocol is Turing complete. I recommend watching this video of an "All-Star Panel" in which Mr Wright confuses all present on the subject, including Nick Szabo. 

In essence, what the applications (EP3449450, EP3449451 and EP3449452, all of which were filed on 28 April 2017) try to do is claim that Bitcoin can be made Turing complete by using locking and unlocking transactions to perform logical operations together with other operations off chain. Mr Wright wrote a paper afterwards about this, titled "A Proof of Turing Completeness in Bitcoin Script", although this has been dissected here and shown to be largely a work of plagiarism. 

The application in question started off, as usual for many European applications, as an international application, publishing in November 2017 as WO 2017/187398 A1. Claim 1 of the application as-published (i.e. not examined yet) reads as follows:

1. A computer-implemented control method comprising the steps of: 

    providing a locking script in a first blockchain transaction, the locking script comprising at least one instruction arranged to:

    process at least one Boolean input; and

    implement the functionality of a logic gate.

The international examiner raised objections that the claimed invention was not new over what was described in Andreas Antonopolous' book "Mastering Bitcoin", first published in 2014, given that locking scripts could involve logical operations such as checking for multiple signatures. 

After the application entered the European regional phase in 2018, nChain's attorney responded to the examiner's objections by filing some amendments and arguments, following which there was a round of examination and some further amendments to the claims. This resulted in the following amended version of claim 1 (where the underlined parts are the attorney's additions):

1. A computer-implemented control method comprising the steps of: 

    providing a locking script in a first blockchain transaction, the locking script comprising at least one instruction arranged to:

    monitoring and searching the blockchain or blockchain network to determine the presence or absence of a further transaction;

    process at least one Boolean input from a further blockchain transaction, the Boolean input provided to execute the locking script of the first transaction; and 

implement the functionality of a logic gate, wherein the logic gate is an XOR gate,

    the method further comprising: providing the further blockchain transaction having an unlocking script;

    processing at least one input signal to provide the at least one Boolean input;

    using the at least one Boolean input to execute the locking and unlocking scripts of the first and further blockchain transactions, providing a computing resource arranged to influence the behaviour of a device or process based upon:

    detection of the further transaction within the blockchain or blockchain network; and/or

    the validity of the further transaction.

In essence what the applicant appeared to be trying to do was to claim that a blockchain transaction could be used to implement an XOR logic gate and control some process as a result of an unlocking transaction being found (think perhaps Ethereum-controlled bike locks). To those familiar with the Bitcoin protocol, all this will appear to be lacking in novelty, given that the original Bitcoin scripting language included various logical functions, including XOR (which was subsequently disabled), and the process of locking and unlocking transactions was the way to interact with the blockchain. As for "influencing the behaviour of a device or process", this could cover all kinds of things that could for example relate to a Bitcoin node checking a received transaction and forwarding it to other nodes or including it in a block to be mined. The invention as a whole therefore appears to be lacking in novelty over what was known well before 2016. 

A lack of novelty, however, is not the main problem with the above claim. If we look more closely at the order of the steps, it is clear from the way claim 1 is now presented that the locking script is the part that contains instructions to "monitoring [sic] and searching [sic] the blockchain or blockchain network to determine the presence or absence of a further transaction". This is not what is disclosed elsewhere in the application, where the monitoring and searching is described as being done by an agent running in parallel to the blockchain network. It should also be evident to anyone familiar with Bitcoin script that this kind of thing is not even possible. As a result, nChain's attorney has (presumably inadvertently) described the invention in a way that was literally not disclosed in the original application, as well as being impossible. Not only that, but the EPO examiner did not spot this error and allowed the application to proceed to grant with this, as well as another error the examiner made, intact. This is not, by the way, up to the proper standard we should expect from the EPO. 

European patent law states that a patent “may not be amended in such a way that it contains subject-matter which extends beyond the content of the application as filed” (Article 123(2) EPC). The feature of monitoring and searching being part of the instructions performed by the locking script was not disclosed in the application as filed, so this adds matter contrary to Article 123(2) EPC. 

Another feature of European patent law is that, once a patent has been granted, its scope cannot be broadened. In the words of Article 123(3) EPC, the patent “may not be amended in such a way as to extend the protection it confers”. Therefore, if the patent has been amended to include a feature that adds subject matter contrary to Article 123(2) EPC, it cannot be removed after grant because this would extend the patent’s protection. 

The result is that the patent is now in an impossible position because the amendment needed that would solve the added matter problem would inevitably extend its scope and cannot therefore be made. Claim 1 of the patent, together with any claims dependent on it, therefore appears to be irretrievably invalid. This, of course, makes the patent incapable of being a real threat to anyone. It will, however, remain on the patent register for as long as the proprietor keeps it alive or, if someone takes action against it, for as long as it survives. 

It happens that someone who has been taking a very active interest in Craig Wright's activities over the past few years, Arthur van Pelt, was keen to have a go at attacking the patent. Arthur has recently, with my help, filed an opposition at the EPO (documents available here and here). Thanks to the generosity of various Bitcoiners, the 840 Euro opposition fee has been paid, with some left over for later. The process is unfortunately not quick and it will take until probably the middle of 2023 before we get any further news and some time in 2024 before a decision is made. The outcome, however, appears in my view to be fairly certain, which is that the patent will very likely be revoked. If nChain manage to rescue anything from it, I will be very surprised. In the meantime, there are other targets we can aim at. If anyone wishes to help, please do get in touch.

Withdrawals: Check and Check Again

Withdrawing a European application can be a useful way of getting back some fees or preventing anything further appearing on the public register for the application. It is, however, a final option and is generally irrevocable once done. It is always therefore slightly scary when a client asks for their European application to be withdrawn, just in case a mistake is made and either the wrong application is withdrawn or the applicant changes their mind or provides conflicting instructions. Withdrawing an application is also final in the sense that, once a withdrawal is requested, the application is no longer pending and a divisional cannot then be filed. 

One possible situation would be if instructions are received to file a divisional and to withdraw the current application so that a refund of the examination fee could be obtained and another attempt made at getting the invention examined. If the divisional was filed the day before the withdrawal request, all would be fine. If, however, the application was withdrawn first the divisional would not be validly filed. This is unfortunately exactly what happened for European application number 15750584.3, leading to decision J 5/19, one that issued last year but which I have only just noticed. The representative requested withdrawal of the application on 3 May 2018 to get a 50% refund of the examination fee, and filed a divisional application a few days later on 8 May 2018. The EPO duly refunded the examination fee and separately informed the representative that the divisional application could not be treated as a divisional because the parent application was no longer pending (Rule 36(1) EPC). The representative then immediately recognised their mistake and filed a request to retract the withdrawal. The examining division rejected this on the grounds that a withdrawal could not be retracted. An appeal was then filed against this decision, with the representative arguing that, because both the withdrawal and the divisional filing appeared on the register at the same time, third parties would not be led to believe that the subject matter had been entirely abandoned, and that they would have reason to suspect that the withdrawal was erroneous because the application had been withdrawn before the divisional was filed. 

The Board of Appeal outlined the conditions to be met for a withdrawal to be corrected, which were that:

(a) the withdrawal did not reflect the true intention of the applicant (existence of a mistake within the meaning of Rule 139 EPC); 

(b) there was no undue delay in seeking its correction (J 04/03, point 9 of the Reasons; J 10/87, point 13 of the Reasons); 

(c) third parties who might have taken note of the withdrawal by inspection of the file would have had reason to suspect that the withdrawal was erroneous (see J 10/08, point 12 of the Reasons; J 2/15, point 13 of the Reasons; J 8/06, point 6 of the Reasons). 

The Board noted that other decisions had added further conditions such as the error being due to an "excusable oversight" (J 4/03), which the Board in this case disagreed with because a correction under Rule 139 EPC did not require this, only that it should be evident that an error had been made and what the correction should be. This did not require the "due care" test that applied in cases of re-establishment. Also, according to the Enlarged Board decision in G 1/12, the conditions to be met did not include any reference to due care. 

To determine whether the withdrawal was due to a mistake, the Board considered that the intention of the applicant and not that of the representative was relevant. The applicant, via their US attorneys, had provided instructions to both file a divisional and to withdraw the application. The withdrawal prior to filing the divisional did not therefore reflect the applicant's true intention. The representative did seek correction without delay by making a request straight after being informed that the divisional was not validly filed. Finally, third parties would be made aware at the same time of the withdrawal and the divisional being filed, which could only lead to the inference of the divisional being validly filed or a mistake having been made. The Board therefore decided that the correction should be allowed, which they did by correcting the date of withdrawal to the day after the divisional was filed.

The lesson to be learned from this case should be fairly obvious, which is to check and double check before withdrawing an application, both to make sure that you have the correct application number and that you are following the client's instructions correctly. Things can be corrected, but going through a few years of appeal proceedings to find out is not a particularly stress-free way of finding this out. 

It Just Might Be A One-Shot Deal

On preparing a request to enter the European regional phase from an international application, one of the key questions to ask is whether the EPO was the international searching authority (ISA). If they were, the written opinion of the ISA, if it raises objections, will need to be addressed. The first opportunity to do this, if international examination was not requested, is in response to the communication under Rules 161 and 162 EPC, which issues shortly after European regional phase entry. Rule 161(1) EPC states that the applicant will be given an "opportunity to comment on the written opinion [...] and, where appropriate, invite him to correct any deficiencies". If no response is filed to such a communication, the application is deemed withdrawn.  

Frank Zappa's lovely white teeth. Source here.

Since it is well known that EPO examiners can get impatient with applicants that are viewed as not progressing their applications towards allowance, if a response at this stage does not move the application forward it can be considered by an examiner as the applicant losing one chance of addressing their objections. I therefore tend to advise applicants that a full response to the written opinion should be made either on regional phase entry or within the six month period given by the Rule 161/162 communication. If any objections remain these can be addressed at the examination stage, where the applicant will always get at least one shot at responding to such objections. 

An important point to note, however, is that examiners are not always reasonable. A safety net in any response filed at the EPO is therefore always advisable in the form of a standing request for oral proceedings under Article 116 EPC in the event the Examining Division (ED) is not minded to allow the application. The ED cannot then issue a decision under Article 97(2) to refuse the application until after issuing a summons to oral proceedings and, if necessary, holding oral proceedings at which a final decision is made. 

A tricky question is how many communications under Article 94(3) EPC can the ED issue before they take things further. Article 94(3) merely states that the applicant shall be invited "as often as necessary" to file observations and amend the application. The interpretation of this can vary widely. Sometimes the ED will string things out for several communications, particularly if the applicant is making efforts each time to resolve the objections. In some cases, however, and particularly if the applicant does not seem to be making much of an effort, the ED can issue a summons after just one communication, or even in some cases as a first communication. Of course, if a standing request for oral proceedings is not on file, this can instead be a decision to refuse, which can take the applicant by surprise. See an earlier post here for a cautionary tale of what can happen if an applicant does not take objections by the ED seriously and does not have such a request on file. 

All this is a lengthy preamble to a decision that has recently been issued by the EPO Boards of Appeal, T 17/22. This relates to an application that was a European regional phase entry where the EPO was the ISA. The application claimed an "oral care composition" (i.e. toothpaste), defined by various ingredients and their properties, including a whitening agent.  The international written opinion had raised various objections covering lack of novelty, inventive step and clarity. 

As usual, the Rule 161 /162 communication invited the applicant to respond to the objections raised and the applicant responded by making some amendments to the claims with supporting arguments. The examiner was not, however, satisfied with the amendments, which included amending claim 1 to incorporate features from claims 2 and 3. The examiner's problem with this, which was set out in a communication issued under Rule 137(4) and Article 94(3) EPC, was that both claims 2 and 3 as originally filed depended only on claim 1. The applicant responded with arguments as to why the combination of features was supported by the application as filed, requesting the examiner "to reconsider the opinion as expressed in the invitation". The examiner clearly still did not agree with the applicant and, instead of issuing an examination report, issued a decision to refuse the application, the grounds for which stated that the applicant "still has not provided a satisfactory basis and/or argument why the introduction of the combined teachings of claims 1-3 should be allowable", and that the applicant "has not requested oral proceedings and had the opportunity to comment and amend more than once and consequently the ED considers that the right to be heard under Article 113(1) is respected". This apparently took the applicant somewhat by surprise, and the decision was appealed, in part on the grounds that the ED had committed a substantial procedural violation because the right to be heard under Article 113(1) EPC had been violated. The applicant this time did request oral proceedings (although, when it comes to the Boards of Appeal, this does not always work).

In considering whether the ED had committed a substantial procedural violation, the Board noted that the expression "as often as necessary" in Article 94(3) required the ED to issue at least one substantive communication under Article 94(3) EPC before issuing a decision to refuse the application, unless a summons to oral proceedings could be issued. This was supported by the Guidelines C-III 5 and the case law in T 305/14. In the present case, the ED had issued two communications before the decision to refuse. Neither communication was, however, considered to be a substantive one under Article 94(3). The first one, under Rules 161 and 162 EPC, was an invitation that allowed the applicant to make voluntary amendments as well as dealing with objections in the international written opinion, while the second one, under Rule 137(4) EPC, was of a formal nature and also therefore not substantive. The ED had, in the Board's opinion, committed a substantial procedural violation by not issuing a communication under Article 94(3) EPC after the applicant's response. The application was therefore remitted to the ED for further prosecution and the appeal fee was reimbursed. 

Although this was, in the end, the correct result for this applicant, it provides another useful example of why it is always a good idea to have a request for oral proceedings on file. You never know when the examiner is going to be unreasonably difficult, and it is not necessarily good practice to find out that you were correct after all by going through a lengthy appeal procedure that could have been avoided by simply adding one line to your response in the first instance. 

A Prolific, and Persistent, Inventor

The name Michael Oluwaseun Bamidele has come up several times over the past few years while I have been reviewing decisions each month from the UK IPO for the CIPA Journal. The latest one (BL O/746/22), which issued just last week, prompted me to take a closer look at what Mr Bamidele has been doing. The latest decision is one of eight issued since 2014 relating to UK patent applications filed and prosecuted by Mr Bamidele. All but one of these decisions, including the latest one, unfortunately resulted in refusal. You might think that this would not demonstrate a particularly good track record for an inventor, but it turns out Mr Bamidele has managed to get five applications granted so far by the IPO out of a total of 16 filed since 2010. The technology fields for these patents are quite diverse, covering cellular communication (GB2476705B), object location (GB2476850B), domestic waste recycling (GB2476937B), holographic projectors (GB2477508B) and a human-powered electrical power supply (GB2523853B). 

According to his LinkedIn profile, Mr Bamidele is the Founder and Product Owner of Dorex Technologies Ltd, which claims on its website to be "a UK based company offering unique solutions and a variety of services". Mr Bamidele introduces his company in the following video. 

Another video below from Mr Bamidele describes a "Green Hover Vehicle Solution", which is described in GB2552190A. The vehicle apparently removes the need for an internal combustion engine by using magnetic repulsion to levitate over a road surface and air turbines to provide propulsion. Figure 1 of this revolutionary patent application is reproduced below. 

Mr Bamidele and his company, which according to him is worth £600 million (based on a £15M investment for 2.5% of the company - see video above), appear to be very much underappreciated. His achievements in obtaining patents at the UK IPO, apparently entirely without the assistance of a patent attorney, are outstanding. I very much look forward to seeing what he achieves in the future. Perhaps we have another Arthur Pedrick in the making, although such a comparison at this early stage in Mr Bamidele's patenting career may be slightly unfair, especially since Mr Bamidele does not appear to own a cat. 

Unanimous A.I.

What is commonly described as "Artificial Intelligence" (AI) is often nothing of the sort. A recent case at the UK IPO is a good example of this, as well as a good example of where the boundary between patentable and unpatentable inventions can lie in the area of computer-implemented inventions. 

UK application GB1805236.5, from Unanimous A.I., initially published as WO 2017/004475 A1, entered the UK national phase in 2018. The claimed invention related to a "collaborative intelligence system for determining a group result in real-time from a group of users presented with a group question". Claim 1 of the application, which in its published form is nearly two pages long, defines a networked computer system in which individual users are sent a group question and submit their input. A group result would then be generated and sent out to the users until a final group result was determined. The final group result then represented the collective decision reached by the users. 

The UK IPO examiner objected that the claimed invention was excluded for being a computer program, a business method or the presentation of information as such, finding that the contribution was "a set of apps that solve the business administrative problem of obtaining a collaborative group result for a group question", which fell within excluded matter under section 1(2) and did not define a technical contribution according to any of the AT&T/CVON signposts. The application then progressed to a hearing, prior to which the applicant submitted some amendments that narrowed the invention to a specific implementation involving pointer locations and vector representations (as in Figure 4 of the application shown here).

The hearing officer considered that the way the invention was defined in terms of the input device went to the heart of the contribution, which was a new mechanism to allow multiple users, in distributed locations, to interact together via vector inputs to their individual devices to collaboratively influence the position of a single, shared pointer and hence to make a collective selection. While not all inventions relating to input devices were inherently technical, in this particular case the hearing officer agreed with the applicant's attorney that the contribution was unquestionably technical and plainly more than a computer program. The application was therefore not excluded and was remitted to the examiner to conclude the examination process. 

What is interesting about this decision is not only that it is one of a very small number that result in the hearing officer finding that an invention is not excluded, but that the decision relies on specifics about an input device for an invention that would otherwise be excluded. The concept behind the invention, which is in any case not new, is one that would not be considered patentable in the UK but the way it is implemented can be. This, of course, limits the scope of the patent to that particular implementation, but does at least give us some useful indications of how it might be possible to overcome excluded matter objections if there is enough subject matter in the application as filed to describe the specific implementations. In this case, the way that users are able to provide their input in the form of a pointer on a screen, which are combined into a collective vector that eventually arrives as a collaborative answer from feeding back other users' inputs, is actually the way that the system appears to work in practice. The question of whether this is actually AI is another thing, of course. 

The Invention is Defined by the Claims?

Every European Patent Attorney knows that EPO examiners have become more strict over the past few years in enforcing the requirement for the description of a patent application to be consistent with the claims. In previous years it was usually enough just to ensure that the summary section said something along the lines of "In accordance with the invention there is provided a widget according to the appended claims". This has gradually progressed to requiring any use of the words "invention" and "embodiment" throughout the entire description to be related to features that are actually within the scope of the claims. For some applications, particularly those coming from US priority applications, such objections can be very time consuming to deal with, typically requiring a detailed search through the description to see what example or embodiment actually falls within the claimed invention. It is  also often difficult to work out with any certainty whether particular examples do in fact fall within the scope of the claims.

The current version of the Guidelines for Examination, at section F-IV, 4.3, requires any inconsistency between the description and the claims to be avoided "if it may throw doubt on the extent of protection and therefore render the claim unclear or unsupported under Article 84, second sentence or, alternatively, render the claim objectionable under Article 84, first sentence". This section, which has been amended a few times over the past years, outlines various ways in which such inconsistencies may arise. 

As a reminder, Article 84 EPC states, "The claims shall define the matter for which protection is sought. They shall be clear and concise and be supported by the description". The first sentence simply defines that the invention is defined by the claims, which is uncontroversial, while the second sentence requires that there be some support from the description. How, though, does this translate into requiring that everything in the description is completely consistent with the claims?  The only decision cited in the Guidelines in support of this is T 1808/06, which states:

"Any disclosure in the description and/or drawings inconsistent with the amended subject-matter should normally be excised. Reference to embodiments no longer covered by amended claims must be deleted, unless these embodiments can reasonably be considered to be useful for highlighting specific aspects of the amended subject-matter. In such a case, the fact that an embodiment is not covered by the claims must be prominently stated" (point 2 of the reasons).

The Board in that case provided no reasoning as to why such amendments were required by Article 84. This single paragraph has nonetheless apparently caused practice at the EPO to become ever more strict. This has led to much frustration among attorneys, who are having to spend more of their time (and consequently their client's money) on combing through specifications to try to comply with the latest strict interpretation of the Guidelines. It does not help that how strict the Guidelines are interpreted tends to vary considerably from examiner to examiner. Complying has now become a game of seeing how far it is necessary to go to persuade a particular examiner that the amendments to the description will get the application over the line to be allowed. 

In 2018, all this clearly became a bit too much for one applicant, who decided to push back and allow their application to be refused by maintaining the words "one embodiment of the invention" in the description relating to features that, in the examiners' view, did not actually fall within the scope of the claims, which were otherwise allowable. The applicant appealed the decision and the Board of Appeal in T 1989/18 has recently decided in their favour. The reasons for the decision are brief and very much to the point, the key parts being points 4 to 7:

4. First and foremost, Article 84 EPC requires that the claims are clear, i.e. that they properly define and delimit the subject-matter for which protection is sought in understandable and unambiguous terms. Claims must be clear in themselves when being read with the normal skills including the knowledge about the prior art, but not including any knowledge derived from the description of the patent application or the amended patent (see e.g. decision T 454/89, Reasons, point 4.1 (vii)).

5. Article 84 EPC only mentions the description in the context of the additional requirement that it must support the claims. Under this requirement, the subject-matter of the claim must be taken from the description, it being inadmissible to claim any subject-matter which is not described. However, when assessing clarity, the description cannot be relied upon to resolve a clarity issue in a claim, nor can it give rise to any such issue if the definition of the subject-matter in a claim is clear per se. In particular, if the claims are clear in themselves and supported by the description, their clarity is not affected if the description contains subject-matter which is not claimed. (emphasis added)

6. When assessing clarity, Article 69 EPC is of no relevance since it is only concerned with the extent of protection conferred as one of the effects of an application or patent (chapter III of the EPC) whenever that extent is to be determined by whoever is competent to do so. Article 69 EPC is not by itself concerned with a requirement of the Convention to be met by an application or patent - in particular, unlike Article 84 EPC it is not concerned with the definition proper of the subject-matter sought to be protected by a claim. Moreover, even if it were possible, for the purpose of Article 84 EPC, to interpret the claims in the light of the description and drawings as provided for in Article 69 EPC in order to establish whether the conditions governing clarity have been satisfied, the board fails to see how that approach could lead to a lack of clarity of the claims (as opposed to a lack of clarity of the description) if the clear terms of the claims did not encompass subject-matter disclosed in the application or patent.

7. Thus, Article 84 EPC cannot serve as a legal basis for the refusal. 

It should be evident that this reasoning is in direct contradiction with the assertion in T 1808/06 (which is not referenced in the decision) regarding what is required to meet the requirements of Article 84 EPC. If the Guidelines were to follow the new decision instead, section F-IV 4.3 would clearly need to be deleted, because there is actually no requirement for any inconsistencies between the description and the claims to be avoided. This would of course be very good news for applicants and their attorneys, who would be able to get granted applications through the EPO system with less hassle and expense. 

Some may say that this could result in European patents being granted with more doubtful claim scope. Personally, however, I cannot see any realistic downside to this. Any arguments over what the claims actually cover can, if necessary, be resolved if they become relevant during post-grant proceedings, in the event a patent becomes important enough to be contested. I would therefore hope that the EPO will take the reasoning in T 1989/18 on board and change the Guidelines, particularly given that the decision the Guidelines currently rely on has no reasoning at all. Failing that, I can see the issue going up to the Enlarged Board fairly soon on a question from the President, which I am sure some will be lobbying for already, or perhaps another Board will want to follow T 1808/06 instead and refer questions that way. Either way, I think we should see some clarity emerging at last within the next year or two.