Global development
The following chapters give an overview of patenting activity as well as the development of scientific publications in the field of GenAI. For patents, all patent family publications relevant for GenAI were identified and analyzed based on patent data from IFI Claims patent database (see Appendix A.1 for a detailed methodology). A patent family is a collection of patent applications covering the same or similar technical content (i.e. the same invention). We conduct our analysis using patent families to count inventions and not several patents corresponding to the same subject matter and filed in different jurisdictions. For the analysis of scientific publications, we used The Lens (Cambia 2024) as a bibliographical analytics tool, which has an extensive coverage of scientific publications (see Appendix A.6 for the main search query).
Advances in deep learning techniques and increasing computing power have spurred the development of GenAI in recent years. The significant advances in GenAI are reflected in the sharp increase in patent activity in the field. Over the last 10 years, the number of published patent families in GenAI models has increased from less than 800 in 2014 to more than 14,000 in 2023. There has been a big surge in patent activity beginning in 2017, with an average annual growth of around 45% since then. This coincides with the introduction of transformers in 2017. In total, the patent search identified 54,358 patent families published in the field of GenAI between 2014 and 2023. Around 89% of this patent dataset (48,398 patent families) were considered active at the end of 2023.
The number of scientific publications has risen even more over the same period, from only around 100 in 2014 to more than 34,000 in 2023. In 2023 particularly, there has been a strong increase in scientific publications. It is likely that the release of very successful and popular GenAI models and tools in 2022 (ChatGPT, Stable Diffusion, LlaMA, etc.) has initiated a new wave of GenAI research. A lot of the most recent research appears to be focused on reducing the size of large generative models, on better controlling the generation process and on exploring various applications and domains.
Patent family publications have also risen in 2023, but not as much as scientific publications. It can be expected though that there will be a similar acceleration for patent family publications in 2024 and 2025, as there is generally an 18-month lag between the filing and publication of new patents (WIPO 2021).
GenAI is still a relatively small part of AI, but is becoming more and more important
When comparing the development of GenAI patent family publications with all AI patent family publications since 2014, it is clear that GenAI is still only a relatively small part of all AI research activity. In 2023, there were 14,080 GenAI patent family publications compared to almost 230,000 AI patent family publications in total. However, since 2017 we can see that the GenAI share of all AI patents has been increasing (from 4.2% in 2017 to 6.1% in 2023) (Figure 11).
Given the massive increase in public interest in GenAI since the ChatGPT launch in November 2022, as well as the explosion of scientific publications in 2023, it is likely that GenAI will also continue to become more and more important in the patent world within the AI field. Given the time lag between the filing and publication of new patents mentioned above, the recent increase in GenAI research activity is likely to become more visible in patent data from 2024 onwards.
Top patent owners
The Chinese companies Tencent, Ping An Insurance Group and Baidu published the most GenAI patent families in the last 10 years (Figure 12). Tencent has launched its own AI chatbot based on its LLM “Hunyuan,” which supports image creation, copywriting and text recognition, among other applications (Tencent 2023). The company uses “Hunyuan” to add AI capabilities to its flagship products such as WeChat to improve the user experience. Ping An Insurance’s AI initiatives focus on GenAI models for underwriting and risk assessment (MarketsandMarkets 2023). Baidu was one of the earliest players in the GenAI space and recently released its latest LLM-based AI chatbot, ERNIE 4.0. Baidu has also developed multiple LLMs for industries such as IT, Transport or Energy (Triolo and Perera 2023).
The Chinese Academy of Sciences (fourth), the Tsinghua University (15th) and the Zhejiang University (16th) are the only research organizations in the top 20 ranking. Alibaba Group (sixth), Bytedance (ninth), BBK Electronics (11th), Netease (12th), Huawei (14th), China Mobile (17th) and State Grid (18th) are other Chinese companies in the top 20.
IBM (fifth), Alphabet/Google (eighth), Microsoft (10th) and Adobe (19th) are the top US companies in GenAI in terms of patent families. IBM has developed a GenAI platform, watsonx, which enables companies to use and customize LLMs with a focus on data security and compliance, as companies can build AI models trained on their own data (Stack Overflow 2023). Alphabet’s AI division DeepMind recently unveiled its latest LLM model, Gemini, which will eventually be integrated into Google’s search engine, advertising products, Chrome browser and other products (Pichai and Hassabis 2023). Microsoft is a major player in GenAI, not only through its large investment in OpenAI, but also through other research activities. For example, Microsoft’s InnerEye project analyses medical scans to detect abnormalities, diagnose diseases and recommend treatment plans (Microsoft 2024).
Rounding off the top 20 are the electronics conglomerate Samsung Electronics (seventh) (Republic of Korea) and the Japanese companies NTT (13th) and Sony Group (20th). Samsung recently announced the development of Samsung Gauss, a GenAI model that can compose emails, summarize documents and translate text, which the company plans to integrate into its mobile phones and smart home appliances (Yoon 2023).
Because of the success of ChatGPT, OpenAI has become a synonym for GenAI in the public eye. However, OpenAI does not appear to have filed any patents for its research activities until the beginning of 2023. An explanation for this might be the non-profit origin of OpenAI. Originally, OpenAI was founded as a non-profit organization that encouraged its researchers to publish and share their work to “digital intelligence in the way that is most likely to benefit humanity as a whole” (Brockmann and Sutskever 2015). OpenAI initially made open-source significant parts of its technology. The company later transitioned from non-profit to a “capped” for-profit model (with a split of OpenAI into the non-profit OpenAI, Inc and the for-profit subsidiary OpenAI Global, LLC with Microsoft as one of the key investors). An alternative explanation might be that OpenAI is opting to retain its IP in the form of trade secrets.
OpenAI seemed first to protect parts of its technology with trade secrets (Keseris and Kovarik 2023). However, six US patents from OpenAI were published in the first quarter of 2024 (three granted and three pending), filed in early 2023, indicating a change of IP strategy and the creation of a patent portfolio.
As this patent landscape report shows, most large tech companies have filed numerous GenAI patents over the past decade to protect future revenues against license assertions. OpenAI’s lack of patents could therefore pose a risk for their IP strategy (LexisNexis 2023) and their recent patenting activity indicates their need for a defensive patent strategy.
A closer look at research activities of research organizations around the globe shows that the Chinese Academy of Sciences has by far the most extensive patent activity. The Chinese institution has published more than 600 patent families since 2014, almost twice as many as second- and third-ranked Tsinghua University and Zhejiang University. The Chinese Academy of Sciences launched its latest LLM model, “Zidong Taichu 2.0,” in the summer of 2023 which supports various data types including video and 3D.
In total, eight of the top 10 and nine of the top 20 research organizations are Chinese.
Apart from Chinese universities, there are four US universities (University of California, Northwestern University, Stanford University, Arizona State University), three research organizations from the Republic of Korea (National Research Council of Science and Technology, Korea Advanced Institute of Science and Technology, Seoul National University), three Japanese research organizations (National Institute of Information and Communications Technology, University of Tokyo, Osaka University) and one Swiss university (Eidgenössische Technische Hochschule Zürich) in the top 20 (Figure 13).
Regarding the top institutions in terms of a count of scientific publications, China and the US dominate with several research organizations in the top 20 (China has eight, the US six). In addition, there are four universities from the UK and one each from Canada and Japan in the top 20 (Figure 14).
Again, the Chinese Academy of Sciences is clearly ahead with more than 1,100 scientific publications since 2010. Tsinghua University and Stanford University follow in positions 2 and 3 with more than 600 scientific publications each. Alphabet/Google (fourth) is the only company that ranks in the top 20 institutions with 556 scientific publications.
However, the raw number of publications as a performance indicator has its limitations, because it does not reflect the impact of the publications. The number of citations that a publication receives is therefore often considered as a more reliable indicator. The citation numbers by institution show higher positions for companies. Alphabet/Google becomes the leading institution by a large margin, and seven other companies are present in the top 20 (Figure 15). The case of OpenAI is also noteworthy. The company has published only 48 articles according to the GenAI corpus (325th institution in term of document count), but these publications received a total of 11,816 citations from other publications of the corpus (position 13 in total). Such impact is outstanding in our study, in particular given that many of OpenAI’s publications are pre-prints and not published in major conferences and journals.
Key locations of inventors
China is at the forefront of global patenting activity in GenAI. China was responsible for more than 38,000 patent family publications between 2014 and 2023, based on the inventor addresses published on patents. Since 2017, China has published more patent families in this field every year than all other countries combined.
With a total of around 6,300 patent families between 2014 and 2023, the US is the second most important research location for GenAI. The Asian countries of the Republic of Korea, Japan and India are also important research locations in GenAI, all ranking in the top 5 countries worldwide (third, fourth and fifth respectively) (Figure 16).
The UK is the leading European location (sixth on a global level), having published 714 patents in the same period. However, Germany is close behind (708 patent families) and has published more GenAI patents than the UK in recent years.
These seven inventor locations account for the majority of patenting activity related to GenAI, contributing around 98 percent of the dataset, with a few contributions from other countries such as Canada, Israel and France.
Figure 17 gives an overview of the current Chinese leadership in GenAI research activities. Between 2014 and 2023, China achieved a world share of almost 70% of all patent family publications in GenAI globally (Y-axis). Even more impressive is that China has also reached very high average growth rates of patent family publications (50% per year, X-axis) in that period among the top inventor locations despite its already large GenAI patent portfolio. Only India had even higher growth rates in GenAI patent family publications (56% per year). The Republic of Korea has also achieved high growth rates in their GenAI patent families since 2014. By contrast, patent families from Japan and the UK have only risen by a little more than 10% per year on average.
Regarding scientific publications, China and the US largely dominate and are at a similar level in terms of publications (Figure 18).
Figure 19 shows that China and the US, despite being comparable in terms of the number of publications, differ strongly in terms of citations. Scientific publications with at least one affiliation in the US received significantly more citations globally than those with at least one Chinese affiliation.
Key filing jurisdictions
An analysis based on filing jurisdictions of GenAI patent families provides an additional perspective. Members of patent families can be filed directly in one or more countries, via national patent offices, via the Patent Cooperation Treaty (PCT) route administered by WIPO or via the European Patent Convention route (EP) administered by the European Patent Office.
Figure 20 shows that China is not only the leading inventor location in terms of GenAI patent families, but also the top country in terms of patent filings. Between 2014 and 2023, more than 40,000 GenAI patent families were filed in China for seeking patent protection. In the US, the number of patent families filed in the country reached more than 10,700 over the last decade.
It is also worth noting that the Patent Cooperation Treaty (PCT) and the European Patent Convention (EP) are relevant routes for GenAI inventors to seek patent protection. Over the last decade, there have been more than 7,700 patent filings under the Patent Cooperation Treaty and more than 3,100 patent filings under the European Patent Convention.
