As innovators and developers step up their activities in response to a global surge in AI adoption, what do they need to know about ‘foundational patents’ and how can they optimise their chances of owning one?
Foundational patents are not a formal legal category, but the term is commonly used to describe early, high-impact patents, that underpin a technological field. An example foundational patient is US 6,285,999 B1, ‘Method for node ranking in a linked database’, which describes Google’s PageRank algorithm as a general-purpose ranking mechanism for linked information, applicable to web-scale search and other link/citation graphs, rather than a narrow, single application.
In the world of AI, foundational patents may relate to model architectures, training or inference techniques, and supporting compute/hardware components, rather than only end-use applications. Beyond making technological milestones in a fast-moving patent landscape, foundational patents can also be commercially significant: if their claims cover a widely adopted technique, the exclusivity rights offered by such a patent can generate significant revenues through licensing. This is because follow-on innovators might look to obtain licences where their products or improvements fall within the scope of the patented claims.
More AI patents globally
The volume of AI-related innovation activity, focused on both core and applied AI, is growing rapidly. The latest data from the UK Intellectual Property Office (UKIPO) confirms that AI-related patents grew from 5.2% to 20.3% between 2014 and 2023. A similar surge in innovation activity is evident in patent filing data globally, most notably in China, the US and Europe, although there are some differences in approach. For example, China is the world-leading filer of AP patents, but many are filed domestically only, whereas AI patents filed in the US tend to generate more citations, which could give them a perceived credibility factor.
It should be noted, however, that these differences reflect prevalent attitudes to innovation and intellectual property (IP) protection in these countries, rather than being specific to AI or computer-related innovation.
The rapid growth in AI patents is indication of strong innovation activity of course, but it also suggests that there is still a significant opportunity for AI innovators and developers to secure a foundational patent. The global AI market is still growing, and patent filings are in the ascendancy, which means innovators and developers can still make their mark and secure a significant stake in what’s to come.
Understanding the AI patent landscape
Understanding when to seek patent protection, and when not, is key to making the most of the AI market opportunity.
As an IP professional, I am often asked ‘How do I know if my innovation is foundational?’ and ‘Should I file now or later?’ An up-to-date, holistic view of the global AI patent landscape can inform such decisions and help innovators to identify technologies that are genuinely ‘foundational’. Filing patents early is usually recommended, even in the case of potentially ‘hidden’ foundational technologies (those that can’t be readily reverse engineered), such as new neural network architectures or training processes. This approach not only blocks competitors from beating you to it and taking advantage of any licensing opportunities that might follow.
An early AI-related foundational patent was EP 3 278 213 B1, which was granted to C3 AI Inc for an enterprise ‘Internet‑of‑Things (IoT) application development platform’ delivered as a Platform‑as‑a‑Service (PaaS). The patent describes a model‑driven architecture and type system that aims to integrate many heterogeneous enterprise and telemetry data sources and enable near real‑time machine learning over very large datasets. Given the early priority date of this patent, and the large number of forward citations the family has attracted, the patent represents a key technological advancement in the deployment of real-time AI systems.
Building on the right foundations
Understanding the patent landscape is crucial for all AI innovators and developers. It could prevent them from investing in a field of research and development that is too close to an existing foundation technology, which could pose a risk of litigation. Alternatively, innovators may wish to target a work-around or obtain a licence agreement to develop incremental improvements for a specific end-use application. Adopting the right innovation and IP strategy at an early stage, could result in outputs that are patentable in their own right.
There is much to consider for AI innovators and developers, but there are still significant opportunities to own the exclusivity rights to ‘foundational’ technologies. For all innovators, the best commercial opportunities are likely to come from understanding the AI patent landscape; spotting gaps and building on the right foundations.
Harry Strange is a partner and patent attorney, specialising in AI, computing and software innovation, at European IP firm, Withers & Rogers.
