AI & Industry
Surge in Generative AI Patents — WIPO Releases Latest Trends Report
AI technologies are advancing at an almost exponential pace. New models and services are being announced one after another, and information that was cutting-edge just a week ago can quickly become outdated.
This acceleration in technological development is also clearly reflected in patent activity. In July 2026, the World Intellectual Property Organization (WIPO) released a report summarizing the latest global patent trends related to generative AI.
More Patents in Two Years Than in the Previous Decade
According to WIPO, the number of published patent families related to generative AI rose from around 14,000 in 2023 to over 37,800 in 2025.
In total, more than 56,000 generative AI-related patent families were published between 2024 and 2025 alone—surpassing the cumulative total recorded over the entire decade from 2014 to 2023. The share of generative AI within all AI-related patent families also increased from 6.1% in 2023 to 8.7% in 2025.
By technology category, patents related to large language models (LLMs) have surged dramatically. In 2025 alone, approximately 14,100 LLM-related patent families were published, far exceeding the roughly 5,200 filings related to generative adversarial networks (GANs), which had previously been widely used in image generation.
Japan Shows the Fastest Growth Among Major Countries
By country, China continues to dominate in absolute numbers. Chinese inventors accounted for more than 43,000 generative AI-related patent families published in just 2024 and 2025 combined.
Japan, however, recorded the highest growth rate among major inventor countries, with a compound annual growth rate of 210% between 2023 and 2025. This also pushed Japan’s global ranking from 4th to 3rd place.
At the company level, SoftBank leads the ranking with around 3,000 generative AI-related patent families, according to WIPO’s dataset. Most of these were filed in 2023 and published in 2025.
That said, this does not mean Japan is leading in absolute terms. China still holds a significant lead in total volume, and the attention Japan is receiving is mainly driven by its rapid recent growth.
Why Does OpenAI Have Fewer Patents?
One of the more interesting findings is that OpenAI—one of the most prominent generative AI companies—has a relatively small patent portfolio compared to other major players.
WIPO reports that OpenAI had only 35 patent filings globally as of the end of 2025. However, this does not necessarily indicate slower technological progress.
Instead, WIPO suggests that OpenAI may rely more heavily on trade secrets and rapid development cycles as its primary competitive advantage, using patents in a more selective and complementary way. Its filings tend to focus on product-level technologies such as multimodal interfaces, code generation, image generation, and text editing.
In other words, patent counts alone are not a reliable measure of a company’s technological strength. Some firms aggressively patent their innovations, while others prioritize secrecy and speed, continuously moving ahead before competitors can catch up.
The Changing Competitive Landscape of AI Development
Until recently, competition in generative AI was largely centered on building larger and more powerful models. However, the focus is now shifting toward other dimensions such as speed, deployment cost, computational efficiency, and usability across different applications.
For example, Anthropic’s recently announced “Claude Opus 5” emphasizes delivering near top-tier performance at significantly lower cost. At the same time, there is a growing trend toward using multiple specialized models—rather than relying on a single monolithic system—depending on the task.
In addition, a design approach known as “orchestration,” where multiple models are combined dynamically, is gaining traction. Sakana AI’s “Sakana Fugu” is one example of this approach, coordinating multiple models depending on the problem at hand.
WIPO also highlights areas such as inference optimization, model efficiency, and AI agents as fields likely to see increased patent activity in the coming years.
What Patent Data Really Tells Us
While patent statistics vary depending on filing routes and procedures, there is typically an 18-month delay between application and publication. As a result, the data reflects not the current state of the generative AI market, but rather R&D efforts and investments made several years ago.
Even so, the fact that 2024–2025 alone surpassed the total number of filings from the previous decade suggests that generative AI has moved beyond a passing trend and is now entering a phase of large-scale intellectual property formation.
Going forward, developments in AI agents, inference models, lightweight architectures, and cost-efficient systems will gradually appear in patent statistics with a time lag. This makes it increasingly important to pay attention not only to the technology itself, but also to the evolving IP strategies surrounding it.