What’s going on the world of AI patenting?
In this article, we’ll look at AI patenting activity and trends over the last five years including the distribution of AI patent grants worldwide, AI patent applications at the IP5 offices and the big AI players.
You don’t need to dig too far into the news nowadays to stumble across AI stories. Just earlier this month, the Royal Swedish Academy of Sciences awarded the 2024 Nobel Prize in Physics to John Hopfield and Geoffrey Hinton of the University of Toronto for their foundational work in machine learning with artificial neural networks, and awarded the 2023 Nobel Prize in Chemistry to Demis Hassabis and John Jumper (along with David Baker) who led the development of DeepMind’s AI model AlphaFold, capable of predicting the structure of proteins. Governmental AI legislation is a constant hot topic. OpenAI has a ubiquitous presence on the technology pages. And new fronts are constantly opening up in the copyright war against current generative AI practices.
Google Trends data for the search string “Artificial Intelligence” show global interest in AI is still on the rise.
Source: Google Trends
But what’s going on the world of AI patenting? Below, we’ll look at AI patenting activity over the last 5 years.
This is a difficult one. There’s no agreed-upon, straightforward way to classify a patent application as an AI patent application. If an innovation makes very incidental use of some AI technology – as many do – is it an “AI patent application”? For simplicity, let’s say that an AI patent application is any application classified by a patent office into at least one of the following CPC groups:
G06N 3/00 Computing arrangements based on biological models
G06N 5/00 Computing arrangements using knowledge-based models
G06N 20/00 Machine learning
Of course, this approach isn’t going to be perfect and will likely miss some applications concerning AI that haven’t been classified as above.
The below figure shows the absolute number of patent applications published by the IP5 patent offices (the EPO, USPTO, CN, KR, JP) since January 2019. This data (and that of all subsequent figures) is accurate up to 20th September 2024; hence, 2024’s data is incomplete.
Source: Lens (https://www.lens.org/)
In this period, there have been a total of over 480,000 AI patent application publications at the IP5 offices.
Perhaps unsurprisingly, China leads the way with a peak of just over 70,000 AI patent applications published in 2022 – this represented a 300% increase relative to 2019’s publications.
At the EPO, there have been a total of over 26,000 AI patent applications published since 1st Jan 2019, with 5,756 AI patent applications published last year (2023).
But how do these numbers compare to all patent application publications?
Source: Lens (https://www.lens.org/)
The above figure shows the share of AI patent applications published by the same IP5 patent offices, relative to the total number of all AI patent application publications.
Since January 2019, there have been just over 15,000,000 patent applications published (in all CPC classes).
Notably, until this year, the US has dominated in terms of the share of AI patent application publications, with around 5% of all patent application publications being AI related. Korea, however, looks to overtake the US by the end of 2024.
Perhaps surprisingly, Japan – despite the generous stance of patenting software and business methods – consistently publishes the fewest AI patent applications (both in terms of absolute numbers, and relative numbers as shown here).
At the EPO, relative AI patent application publication numbers rose steadily to around 3% in 2021 and have remained at this level ever since.
Published patent application data typically gives more information about technological activity than granted patent data because of factors including national/regional patentability criteria and the patenting strategies of applicants who may file more applications than they ever intend to pursue to grant. Nonetheless, it’s interesting to see the distribution of AI patent grants worldwide.
China’s AI industry is huge. The State Council has a list of “national AI teams“, including fifteen China-based companies, including Baidu, Tencent, and Alibaba; each company is expected to lead the development of a designated AI sector in China, such as facial recognition, software/hardware, and speech recognition. It’s not too surprising then than the Chinese AI patent grant numbers are so high. China rapidly overtook the US in 2021 and topped 35,000 granted AI patents last year (recall there were 67,000 AI patent publications in China last year).
Source: Lens (https://www.lens.org/)
Perhaps reflective of the relatively strict stance on CIIs, the EPO consistently grants the fewest AI patents amongst the IP5 offices: last year, there were just 1,380 AI patent grants.
Again, let’s look at how these numbers compare to all patent grants.
Source: Lens (https://www.lens.org/)
Despite the high absolute number of CN AI patent grants, the US still consistently leads in terms of the relative number of AI patent grants: in 2023, 5.3% of all US patent grants were AI patents.
In comparison, in the same year, just 1.3% of all EP patent grants were AI patents.
This data perhaps corroborates that commonly held belief that it is trickier to obtain software patents in Europe than it is in the US.
This table shows the most active applicants in terms of the number of patent applications published by the EPO.
In green, on the left, are the most active applicants, regardless of application subject-matter. You can see Huawei lead with over 21,000 applications published from January 2019 to today.
In grey, on the right, are the most active applicants for AI patent applications. Samsung are the most active in the AI patenting space, with over 1,000 published AI applications.
In bold are those applicants who appear in both tables. The top 5 AI EP applicants (on the right) are also top 10 general EP applicants (on the left).
This table shows the most active applicants in terms of the number of patent applications published by any of the IP5 patent offices.
As well as being the most active AI applicant in Europe, Samsung are the most active applicant for AI and for all subject-matter areas in the IP5 offices.
Just as the most active AI applicants in Europe were also some of the most active general applicants in Europe, the top 3 AI applicants worldwide are also some of the most active general applicants worldwide.
But where are these applicants all from?
Looking only at EPO application publications (it seems that the IP5 offices don’t capture applicant nationality data in the same way), we see that US applicants are the biggest source of EP AI publications.
Source: Lens (https://www.lens.org/)
Shown above are the top 5 EPO AI applicant sources over 2019 to today: the US, Germany, China, Japan, and Korea (which interestingly align nicely with the IP5 offices).
Plotted are the absolute numbers for EP AI patent application publications. For example: in 2022, from a grand total of just under 5,700 EP AI publications, just over 4,000 included an applicant from one of the 5 listed countries. In 2022, there were 2,225 published EP applications with a US applicant.
Below, we have the publication numbers for EP AI patent applications for the same 5 applicant sources, as a percentage of all EP AI patent application publications.
Source: Lens (https://www.lens.org/)
Interestingly, there doesn’t seem to be any significant change in behaviour from any of the top 5 applicant sources – the US consistently accounts for around 37% of EP AI publications; the top 5 sources steadily account for just over 70% of EP AI publications.
This is in spite of the previously shown rapid increase in AI patent application publications at the CNIPA. If we presume that most applicants at the CNIPA are Chinese, it seems like Chinese AI applicants have not adapted their EP filing strategies in line with the increase in AI activity at home.
The number of AI patent application publications have shot up worldwide and we’re seeing more and more of these applications proceed to grant. At the EPO, we’re seeing masses of patenting activity from all over the globe.
It doesn’t look like interest in AI is going away anytime soon. As digital technologies continuously develop, resourceful inventors are crafting novel AI models and finding exciting ways to put existing AI tools to good use. We’re excited to see where AI takes us next.
This is for general information only and does not constitute legal advice. Should you require advice on this or any other topic then please contact hlk@hlk-ip.com or your usual HLK advisor.
Keep up-to-date with the latest IP insights and updates as well as upcoming webinars and seminars via HLK’s LinkedIn page, or simply subscribe to our updates.
What are latest trends and best practices shaping the world of materials and semiconductor patents?
Register now