02 October 2020

Securing intellectual property protection in machine learning techniques for fintech businesses

Fintech places particular demands on high frequency computing (HPC) that requires significant investment into computing resources such as CPU and GPU, in addition to the requirements of data scientists’ and quants’ time to maintain a complex and ever-evolving code base. New, more effective solutions are required as sophistication of regulations and increased risks grow.

Machine learning (ML) is also increasing in prevalence. In so-called supervised learning, ML models are trained on input data sets that have known outcomes; they are then able to ‘learn’ from these and predict outcomes for new input data sets.

The intellectual property (IP) in such techniques is of enormous value. Over the last few years, there have been significant changes in the legal landscape, both in terms of trade secrets protection and what is eligible for patent protection in major jurisdictions. More changes are coming. This article explains some of the challenges surrounding IP in this area and gives practical direction in how to approach the issue.

Changes in the legal landscape – risks and mitigations

The main patent granting authority for Europe (the EPO) is preparing to issue a landmark decision on the granting of patents for the computational techniques used for ML and the modelling of complex systems. There is justifiable speculation that that decision will herald a more liberal approach to the granting of patents in this area – many of which may have clear relevance to the fintech sector. This is important and financial institutions should take note.

Patent infringement represents a very significant hazard. For an infringer in financial services, the cost of paying damages to the patent holder could be just one of the problems they face because a patent holder may obtain an injunction overnight to prevent use of a patented technology. The infringer may thus be forced to withdraw service from a client immediately. If such sudden disruption of service causes consequential losses for the client, the infringer may be liable to their client for those losses. This is to say nothing of the damage to the client relationship, and the not insignificant financial cost and reputational damage of being found to have infringed a patent.

Patents are unique amongst IP rights in the fintech space. Confidential information and trade secrets in Europe only provide a private contractual right against those who have agreed to confidentiality. Copyright is a little broader, but direct or indirect copying must be proven if copyright is to be enforced. Neither copyright nor trade secrets offer any protection outside these limited bounds. Patents on the other hand can be enforced against anyone who uses a technology, regardless if they learned it honestly or created it independently themselves.

This poses a problem: how can fintech innovators protect themselves against patent risk?

Help is at hand: the strategies that have been used in other fields of engineering for decades can be applied directly to fintech. Innovators must conduct sensible patents searches to identify the most likely risks. If found early, problem patents can be opposed at the EPO before they come into force, because defending against them later is at least two orders of magnitude costlier and riskier. This cannot catch everything however, so the most widely used strategy for controlling risk from third party patents is to secure patents of one’s own. The aim of this ‘defensive strategy’ is that one’s own patents present a risk to the competition, or at least cover desirable technology. A competitor who knows they might themselves infringe a patent is unlikely to assert their own patents against the holder of that patent. This strategy has kept the peace in the telecoms and electronics industries for generations, and a leading patents judge once described it as “the only rational way to use the patent system”. What makes the ‘smartphone wars’ newsworthy is that they buck this decades’ long trend.

Case study

Dmitri Golubentsev is the founder of Matlogica, a young fintech startup offering a niche proposition for large financial organisations focused on improving risk management capabilities in terms of performance and cost effectiveness. As a challenger startup, Dmitri says he has proven that the technology their competitors do not believe is possible brings demonstrated and significant performance improvements.

Technology designed to improve performance of applications built using high-level, object-oriented languages has always been challenging to protect with patents. However, Matlogica’s approach works by breaking down and rebuilding the primal code into a recorded function, which is thread safe and can be safely executed on multicore systems. This provides a technical benefit which is the basis of patentable technology.

A complete strategy

There is more to life than patents. Any business operates in a landscape of commercial relationships. The IP and trade secret issues which arise in that landscape just add a further dimension to the existing considerations
in managing those relationships. Other factors may be more or less important at different stages, but taking careful and measured steps at the right time can increase leverage and achieve significant market impact. These steps can be as simple as just ensuring sensitive information is handled properly; ensuring that agreements with development partners are properly drafted and address the key IP issues; or identifying patentable inventions and choosing when to protect them.

Asked to comment for this article, Dmitri said: “Although we are now publicly recognised by leading experts as a breakthrough technology, it is natural to expect some friction during our business’ first steps. By investing in IP, we are not only reducing our barrier for entry, but also increasing chances for partnerships with established vendors. Our patent and IP position allows us to approach a wide range of potential clients, partners and competitors without risk of the technology being easily copied.”

It is common for fintech businesses to overlook these questions, or assume that technology is not patentable, or that questions of ownership can be dealt with once the value in a technology has been demonstrated. Experience shows that this is not the right approach: the value of simple steps done early cannot be overstated.

A version of this article was originally published in The Patent Magazine in October 2020, and a version of this article featured in Fintech Bulletin in September 2020.