The increasing digitisation of the energy industry depends in part on two things: data and AI. Whether it is management of a smart city, monitoring a distribution grid or balancing supply from variable renewables to match grid demands, data fed back to central control is a key digital asset for modern energy companies. Likewise, AI built into control systems to augment or replace human control is on the rise.
Is it possible to protect these key assets with IP rights? IP rights generally have developed alongside more traditional technologies and businesses. However, this article explores the challenges and opportunities of IP protection in the digitisation of the energy industry.
Protecting and exploiting data
Data may come in many shapes and sizes. If the data is about customers or employees, issues of data protection (in particular under the GDPR) will arise, but that is not the subject of this article. Data in general is gathered into a central repository in such a way that the whole collection of data becomes much more valuable than each piece was: the performance data collected across an entire sub-grid rather than just on a single turbine, for example, may allow richer analysis.
In part this fact is reflected in the legal tools available to protect commercially valuable data. There are four main legal tools that may come into play, each with its own role:
- Confidential information/trade secrets;
- Contractual restrictions;
- Copyright; and
- Database right.
Protecting data as confidential in the UK requires that the data have a 'quality of confidence' and, generally, steps be taken to ensure the data is not publically accessible. In practice where data is e.g. performance data about a distribution system it is likely that the data will not in fact be readily accessible to outsiders but suitable measures should be put in place to ensure this is the case in order to show the information is regarded as confidential. Restrictions will need to be put in place on employees' use of the data and on any third party with whom it may be desirable to share the data. The requirement for "reasonable steps" to be taken to protect confidential information in order to satisfy the recently introduced definition of a trade secret under the EU Trade Secrets Directive (which has now been implemented in the UK) also warrants this approach.
These restrictions overlap with the next tool, contracts, which should be used to embody these restrictions. The contract should make clear that the data set is confidential and what may or may not be done with it and by whom. If the data should be returned from a third party after their permitted use of it ceases, some thought must be given to how this can be practicably achieved.
Contracts can also be used to license the rights that may arise under copyright or database rights if it is desirable to share the data with a third party. However, the application of these rights here is not straightforward.
Copyright protects defined categories of 'works': literary works, musical works, broadcasts, etc. It protects the copying of or dissemination to the public of these works. Within this framework, it is possible that a collection of data may in some circumstances qualify as a literary work in the form of a table, compilation or database. A database is the most likely to be relevant here but in order to qualify for copyright protection the database in question must be the intellectual creation of an author (or group of authors) by reason of their selection and arrangement of the materials in the database: it is questionable whether this criterion can be fulfilled with a collection of, for example, technical performance data where the selection and arrangement may be minimal. However, if there has been intellectual investment in deciding what materials should go into the database and how they are to be arranged, protection may be available. Note, however, that this copyright does not protect the individual items of material in the database which may be another key limitation in this context.
Besides the possibility of copyright, some databases may also be protected by database right. Confusingly, this 'database right' is a different legal right than the database flavour of copyright. Note also that this is an EU-derived right and may be subject to change following Brexit. A database for this purpose is a collection of individual data items arranged in a systematic or methodical way and individually accessible. In order to qualify for protection by a database right, the database must additionally have been the result of substantial investment in obtaining, verifying or presenting its content. Importantly, this investment must be in those specific activities rather than in creating the individual data items. And, like the copyright that exists in some databases, the database right does not protect the individual data items. Instead, the database right provides its owner with the ability to prevent the extraction and/or re-utilisation of a substantial part of its content.
At one level, AI systems are just like any other computer software. Thus the same IP rights may be used for their protection: the source code for the AI is protected by copyright. The source code can also be kept confidential such that it becomes a trade secret.
If it is new and inventive, an AI-based system may be capable of protection by filing a patent. However, in Europe computer programmes "as such" are not patentable: the patent applicant must demonstrate the invention has some real-world 'technical effect', e.g. as a control system. In general, there is no reason why this cannot be done for a system involving AI and control systems involving AI control provide a clear hypothetical example of this. However, AI systems concerned principally with analysis present a much more challenging area for potential patent filings. Not only are computer programmes "as such" excluded from patenting but so too are mental acts and mathematical methods. Thus one can envisage that patent applications in an area such as AI systems for demand/supply balance control systems might be capable of patent protection but the invention must be expressed in terms of what the system achieves and not how its analysis is performed. The latter will also need to be explained but will not be the subject of protection where not alloyed to the system's overall effects.
The two together
So far so good. But there is a significant interaction between data sets and AI. AI is often used to deal with problems with large data sets where humans struggle or take too long due to the amount of data. Machine learning AI systems of the kind used here must usually be trained by supplying to them appropriate example or simulated data sets for them to learn on before they are unleashed on the real world. This raises a number of points, such as:
- First, training data sets may be very valuable. The AI itself may be generic but comes into its own when trained for the particular application. Or the AI might be subject to a patent application which discloses its basic mode of operation but much of the value of the AI system may lie in using an appropriately trained AI system. This may give rise to opportunities such as the co-licensing of a patent and one or more accompanying training data sets in a model not dissimilar to that used in chemical and pharmaceutical industries where important 'know-how' accompanying a patent licence can be a very valuable part of such a transaction.
- Second, in many cases the holder of the data set will not be the developer of the AI system but someone who wants to use an AI system developed by another. For a productive collaboration, both parties should ensure they consider how they intend to protect their asset before sharing with the other.
- Third, the training data set will generally have been carefully selected. Often a considerable amount of effort will have been expended in optimising the content of it: this may assist in showing that it is protectable by database right in particular, copyright and (if kept secret) as confidential information/a trade secret having a necessary quality of confidence.
Some consideration must also be given to the reverse situation. An AI control system may gather a valuable data set. Of course this can be done in such a way that the data is kept confidential and contractual restrictions can be imposed on employees, sub-contractors, consultants, etc, in the usual way.
But, can the data set also qualify for protection by copyright or as a database right? If so, who owns that right in various scenarios, such as if the AI system is supplied by a third party? Interestingly, copyright law in the UK provides for the situation of 'computer-generated works' of which these might be an example. The person(s) who made the arrangements for the computer to generate the work is considered in law to be the author of the work, although this area has not been tested for AI systems. However, the same is not true of the database right which does not have this provision: it may well be argued that the same principle should apply since, at least in the right circumstances, the investment in the AI system could be said to be the investment in the obtaining, verifying and presenting of the data items in the database. However, this has not been established and clearly opposing arguments are readily available.
As regards confidentiality, if the AI system is set up such that the data is only accessible by certain people there seems to be no reason why the involvement of the AI system should undermine the necessary quality of confidence needed to acquire protection for that information. However, one can imagine a situation in which this position could be undermined in circumstances where a less sophisticated AI system that gathers very mundane data also gathers data of greater potential value but does not itself discriminate between the two.
In the right circumstances AI systems and data assets can be protected by powerful legal tools. However, consideration should be given to which legal tools are best for the particular situation and the systems then implemented in a way that will best ensure protection is acquired.
Our lawyers at Bird & Bird can advise on putting the above into practice to help businesses achieve protection of their digital assets.