The World Intellectual Property Organization released a new report showing generative AI patents nearly tripled from 2023 to 2025, skyrocketing from about 14,000 to over 37,800. This sharp rise in just two years surpasses the total patents filed in the previous decade combined.
China dominates this space, boasting more than 38,000 generative AI inventions between 2014 and 2023 roughly six times the number filed in the US. Major Chinese players like SoftBank, Tencent, and Baidu hold a significant share of these patents, illustrating a strong concentration of power in the region.
The United States and Japan are increasing their patent activity, but most filings remain domestic, especially in China, which suggests potential barriers for international innovation.
One notable shift revealed by the data is the rise of large language models (LLMs) overtaking generative adversarial networks (GANs) as the most patented AI technology. While image and video applications still lead in use cases, the core AI architectures steering these innovations are now focused on language models, driven by advances following the 2017 transformer paper.
The report highlights that generative AI patents grew from 4.2% of all AI patents in 2017 to an expected 8.7% in 2025, reflecting the explosive commercial interest sparked by LLMs after 2022.
This surge creates significant challenges for decentralized AI and crypto projects that rely on open-source models and token-based incentives. With companies like SoftBank and Tencent owning thousands of patents, smaller community projects might encounter increased legal risks and obstacles.
As decentralized AI protocols aim to build alternatives to corporate-controlled systems, the tightening patent environment could stall innovation and complicate development.
