Months before ChatGPT launched, Marks & Clerk's patent data pointed to the generative AI wave. Five years into tracking AI filings at the European Patent Office, the firm's next report, due to be published on 25 November, says the frontier has moved again, this time into hardware.
The data saw it coming
In 2022, months before ChatGPT reached the public, our annual report on AI patent filings at the European Patent Office (EPO) named the next wave of innovation. Natural language processing, multimodal architectures, GPT-3, and AI-generated code and imagery were all flagged as the areas to watch.
We didn't have a crystal ball. We had the filing data, and we had attorneys who knew how to read it. Patent applications are filed long before products reach the market, which makes them one of the earliest signals of where technology is heading.
That principle has guided our analysis since 2021. As we prepare to launch our sixth report, it is worth showing how well it has worked, and where it points next.
Five reports, five turning points
What began in 2021 as a software-focused study has grown into a firm-wide annual effort. Attorneys from our energy, med-tech and quantum practices have contributed alongside our software specialists, because AI no longer sits in one technical field. Each edition has captured a distinct moment in AI's journey from niche technology to industrial infrastructure:
| Year | Focus | What the data showed |
| 2025 | The quantum frontier | An emerging patent race at the meeting point of AI and quantum computing, with a lower allowance rate for quantum-AI filings than we expected |
| 2024 | AI and the energy transition | Machine learning moving into grids, renewables and environmental technology, with significant untapped IP potential |
| 2023 | The generative wave | Examination and legal friction around generative models, deepfakes and AI inventorship after the DABUS cases |
| 2022 | Commercial maturation | South Korea's rise as an AI patent powerhouse, sustained growth in computer vision, and more EPO oppositions as portfolios gained commercial value |
| 2021 | Mapping the baseline | A surge in filing volumes, shifting geographic distribution, and the EPO's evolving approach to mathematical methods and technical effect |
Our predictions have held up. In that first report, we said that:
- EPO filing volumes would keep climbing steeply
- China and South Korea would aggressively expand their European filing footprints
- Patent offices would refuse to recognise AI as a legal inventor
- The EPO's "technical character" requirement would become the defining battleground for commercially valuable protection
All four have come to pass. Perhaps most importantly for innovators, the line between an unpatentable algorithm and a protectable technical invention now underpins almost every AI filing strategy in Europe.
We won't pretend we foresaw everything. Nobody predicted how fast generative AI would explode into public life. And our 2025 finding on quantum-AI allowance rates genuinely surprised us, a reminder that letting the data lead, rather than our assumptions, is what makes the analysis worth reading.
AI gets physical
The clue was already in our 2022 report. Alongside the generative wave, we flagged a surge in industrial automation and robotics in which AI would leave the cloud and enter physical systems. Four years on, that thread has become the main story.
Our sixth report, launching next week, turns to the crossover of AI and hardware. It examines where the next wave of filings is building, across robotics, custom silicon, neuromorphic architectures and memory.
The stakes are particularly high within this sector. As algorithms and the chips that run them evolve together, the most valuable AI inventions increasingly sit in the machine rather than only in the model. Protecting them calls for a different strategy from the one that served pure software, and the filing data already shows who is moving first.
Why reading the data takes more than data
Anyone can count patent filings but turning them into reliable predictions takes attorneys who prosecute AI applications at the EPO every day, and who understand the technology behind them, from power grids to qubits to semiconductors.
That breadth is why our forecasts have held up for five years. It is also why the businesses building AI's physical future, from robotics start-ups to chip designers, need advisers who can see both the technology and the legal landscape it is moving into.
If the last five years are any guide, the patterns in our next report will be tomorrow's headlines. Watch this space.
Subscribe to receive more articles like this here.


