Table of Contents
Few technologies have been patented as fast as this one. Patents on artificial intelligence have gone from a niche filing category to one of the most contested arenas in the global IP system, and the numbers behind that shift tell you where R&D capital is really moving. This guide pulls together the primary data – from WIPO, the EPO and Stanford’s AI Index – to answer the questions that matter for strategy: how many AI patents exist, who owns them, which techniques are growing fastest, and whether an AI can be named as an inventor at all.
How Many Patents on Artificial Intelligence Exist?
The clearest headcount comes from WIPO’s 2024 Patent Landscape Report on generative AI, which found more than 54,000 generative-AI patent families published between 2014 and 2023. That is not a flat total – it is a curve. Annual publications rose from just 733 families in 2014, and WIPO reports that generative-AI patenting has grown by more than 800% since the transformer architecture was introduced in 2017.
Two points matter for anyone reading these figures. First, a patent family is counted once across all the jurisdictions it is filed in, so 54,000 families represents a far larger raw filing count. Second, the acceleration is recent: the largest share of activity landed in the last two years of the window, which is exactly the signal a strategist wants – the field is still bending upward, not plateauing.
- 54,000+ generative-AI patent families published 2014-2023 (WIPO)
- 733 families in 2014, rising sharply after the 2017 transformer
- 800%+ growth in generative-AI patenting since 2017
- Scientific publications on GenAI reached 34,000+ in 2023 alone
Who Owns AI Patents: China, the US and the Top Applicants
Ownership of patents on artificial intelligence is strikingly concentrated by country. WIPO credits China with more than 38,000 generative-AI patent families over 2014-2023 – and since 2017 China has published more families in the field each year than the rest of the world combined. The United States is a distant second with roughly 6,300 families, followed by the Republic of Korea, Japan and India.
At the applicant level the concentration is just as visible. WIPO’s top generative-AI filers are Tencent, Ping An Insurance Group, Baidu, the Chinese Academy of Sciences and IBM. What that list tells you is where the core-architecture positions are being staked – and, by omission, where they are not. Adjacent layers such as AI governance, evaluation and compliance remain comparatively thin in the patent record, which is where later entrants often find defensible white space.
For a company planning its own filing programme, this is the difference between walking into a crowded room and finding an empty one. Reading assignee concentration before you file tells you whether you are about to spend budget contesting Tencent and Baidu head-on, or building in a segment almost no one has claimed.
Which AI Technologies Are Filing Fastest
Aggregate totals hide the most useful signal. Broken down by technique, the WIPO data shows sharply different growth rates:
- Large language models (LLMs): patent families grew from 53 in 2020 to 881 in 2023
- Diffusion models: from 18 families in 2020 to 441 in 2023
- Generative adversarial networks (GANs): about 9,700 families over 2014-2023
The LLM and diffusion curves are the story of the last three years – near-vertical filing growth that mirrors the products now reaching the market. GANs, by contrast, represent an earlier and larger accumulated base whose growth has cooled. Segment-level curves like these are why patents on artificial intelligence are such a useful market indicator: they show which techniques R&D teams are committing capital to well before that commitment shows up as revenue.
AI Patents in Europe and the Investment Behind Them
The filing appetite is not confined to Asia. The European Patent Office logged a record 201,974 patent applications in 2025, and computer technology – the field that captures most AI inventions – grew 6.1% on the back of AI advances, according to the EPO’s Technology Dashboard 2025.
The money confirms what the filings imply. Stanford’s 2025 AI Index reports that generative AI attracted $33.9 billion in global private investment in 2024, up 18.7% year on year and now more than a fifth of all AI private investment. Total corporate AI investment reached $252.3 billion, with US private AI investment alone at $109.1 billion. Patents follow that capital – which is why the filing curves and the investment curves tend to move together.
For product and corporate-development teams, the practical takeaway is that AI is now a fully funded, globally contested patent field. Deciding where to build – and where to file – without reading both the filing data and the investment data means flying blind in the single most competitive technology arena of the decade.
Can You Actually Patent AI? Inventorship in 2026
A common confusion is whether the AI itself can be an inventor. The answer, in the United States, is settled: no. In late November 2025 the USPTO issued revised guidance on AI-assisted inventions, implementing Executive Order 14179 and rescinding the earlier Biden-era approach.
The guidance reaffirms that only a natural person can be named as an inventor. An AI system – generative or otherwise – is treated as a tool “analogous to laboratory equipment, computer software, research databases,” not as a co-inventor. This tracks the Federal Circuit’s decision in Thaler v. Vidal. Practically, it restores the traditional “conception” test: inventorship turns on whether a natural person formed a definite and permanent idea of the complete invention, and any priority claim must include at least one overlapping human inventor.
So the 54,000+ patents on artificial intelligence counted above all name human inventors – even where AI tools helped generate the invention. That is a crucial distinction for any team using generative tools in R&D: keep clear records of the human conception behind each claimed invention, because that is what the USPTO now examines.
What AI Patent Data Means for Market Strategy
Read the right way, patents on artificial intelligence are a leading indicator, not a scoreboard. Because a family is typically published 18 months after priority and the product ships later still, a rising segment curve is a paid-for commitment to a direction that has not yet reached the market. That is a gift for market sizing and opportunity analysis.
We put this to work directly. In our artificial intelligence market sizing case study, segment-level filing curves and public investment data were used to build a defensible, bottom-up TAM/SAM/SOM for a generative-AI fundraise – the kind of number that survives investor diligence because every input is sourced. That method is the core of our market sizing and opportunity analysis service.
Whether you are raising capital, planning a build, or deciding where to file, the same discipline applies: define segments against filed technology, read the growth per segment rather than in aggregate, and tie the opportunity to the patent position that protects it.
How PerspireIP Can Help
At PerspireIP, our team helps innovators and businesses protect what they build. Whether you need a patent or trademark search, prior-art analysis, or an IP strategy tailored to your goals, we can help. Contact our team to discuss your next step.
Frequently Asked Questions
How many patents on artificial intelligence are there?
WIPO’s 2024 Patent Landscape Report counted more than 54,000 generative-AI patent families published between 2014 and 2023, up from just 733 families in 2014. Generative-AI patenting has grown by over 800% since the transformer architecture appeared in 2017.
Which country files the most AI patents?
China, by a wide margin. WIPO credits China with more than 38,000 generative-AI patent families over 2014-2023, ahead of the United States at roughly 6,300, then the Republic of Korea, Japan and India. Since 2017 China has out-filed the rest of the world combined in this field.
Can an AI be named as an inventor on a patent?
Not in the United States. Revised USPTO guidance issued in late November 2025, implementing Executive Order 14179, reaffirms that only a natural person can be an inventor and treats an AI system as a tool. This tracks the Federal Circuit’s ruling in Thaler v. Vidal.
Which AI technologies are growing fastest in patents?
Large language model families grew from 53 in 2020 to 881 in 2023, and diffusion-model families from 18 to 441 over the same period. Generative adversarial networks make up a larger accumulated base of about 9,700 families, but their growth has cooled relative to LLMs and diffusion.
Why do AI patent filings matter for market sizing?
Because companies file before they ship. A patent family is usually published about 18 months after priority, so a rising segment filing curve signals where R&D capital is committed well before it becomes revenue – making it a strong leading indicator for bottom-up market sizing.