The IP in France's AI surge

Samuel Deschamps at Santarelli reviews the surge of AI start-ups and scale-ups that is happening in France, highlighting the cultures for IP and open source that are being adopted within this ecosystem

---

Share

Samuel Deschamps, Santarelli

---

In In the past few years, artificial intelligence has hit the news, as companies such as Alphabet (the parent company of Google), Microsoft, Meta or Nvidia invest heavily in it. In August 2023, Goldman Sachs estimated that AI investment could be approaching $200 billion globally in 2025, most of which is now private.

Smaller firms, such as Mistral AI, Hugging Face and Nabla, have also taken a good position with respect to innovation in AI. In terms of IP, it appears that they do not specifically rely on patenting innovations, mostly developing proprietary algorithms and managing proprietary data instead.

Mistral AI

French newcomer Mistral AI, founded in April 2023 by veterans of Meta Platforms and Google DeepMind, is making waves by championing open-source technology, developing freely available large language models as an alternative to closed-off systems. They secured €385 million in funding by October 2023 and their valuation skyrocketed to over $2 billion by December 2023. In June 2024, they secured an additional €600 million, propelling their valuation to a staggering €5.8 billion. This latest funding round was led by venture capital giant General Catalyst with the participation of existing investors.

Hugging Face

Hugging Face, a French-American company with roots in New York, creates tools for building applications with machine learning (ML). Its claim to fame is the Transformers library, ideal for natural language processing, and a platform for sharing ML models and datasets. Originally founded in Paris in 2016, the company started with a chatbot app for teenagers, but quickly pivoted to become an ML platform. Its dedication to open-source technology is evident in its contributions to large language models like BLOOM, a multilingual powerhouse with 176 billion parameters.

Hugging Face’s growth has been impressive. It has secured significant funding rounds, reaching a $4.5 billion valuation by August 2023. Partnerships with tech giants like Amazon Web Services (AWS) and collaborations with Meta to accelerate European AI development showcase their commitment to making AI accessible and fostering innovation.

Nabla

Paris-based start-up Nabla is an ambient AI assistant, designed to take the burden off healthcare professionals and improve patient care. Nabla listens in on your consultations and uses its cutting-edge AI to automatically generate clear, detailed clinical notes that seamlessly integrate with your electronic health record. Developed by top AI researchers alongside practising clinicians, Nabla works across specialties and settings, offering numerous options for customisation to fit any workflow. It is already trusted by over 30,000 clinicians and used by over 40 organisations, making it the most widely implemented ambient AI in healthcare. Its record for constant innovation with the aim of becoming the most intuitive and reliable AI partner for all clinical needs has secured this French start-up $24.1 million in series B funding.

On average, according to his calculations, universities capture 16 percent of the value they help create through licensing revenues or equity stakes in the start-ups that their research spawns. Within those figures, it appears that some researchers and universities are much better able to commercialise their discoveries than others.

He has isolated two factors that shape this commercial landscape. One is the degree to which an academic team is interdisciplinary. The second is the presence of a star commercialiser.

‘Interestingly enough, it’s not the star academic that is correlated with commercialisation,’ says Hsu. ‘It’s the star commercialiser, those with substantial prior experience bringing products to market via start-up formation.’

France: thriving hub for AI in Europe

Inspired by ventures like Mistral AI, Hugging Face and Nabla, France is rapidly becoming a powerhouse for AI in Europe. On 21 May 2024, the French government signalled it as a strategic priority by announcing a €400 million investment fund, Here’s a look at how its AI ecosystem is now evolving.

• Start-up boom: with over 600 AI start-ups (a 24 percent increase from 2021 to 2023), France boasts a vibrant entrepreneurial scene. Notably, half of these start-ups are already profitable or on track to be within the next three years.

• Generative AI leadership: France takes the lead in generative AI with 76 start-ups specialising in creating realistic sounds, text, videos and images.

• Investment magnet: AI companies in France secured €3.2 billion in funding in 2022, making it a top destination for investors.

• Global AI talent: France is the top choice in Europe for foreign investment projects in AI, highlighting its attractions for AI talent.

• Home to AI giants: major players like Google, Meta, Hugging Face, Dataiku, Tata and Accenture have chosen France as their AI research and decision centre.

• Cutting-edge research: France is at the forefront of AI research, having developed some of the world’s most powerful models, including Llama 2, Llama 3 and Mistral Large.

France’s commitment to AI extends beyond current achievements. The Institut Polytechnique de Paris, a prestigious research institution, recently received €70 million to create a world-class AI cluster. This initiative aims to:

• Foster groundbreaking research and innovation in AI.

• Train the next generation of AI experts.

• Develop practical AI applications for societal benefit.

This investment aligns with the France 2030 plan that aims to establish a network of ten internationally recognized AI clusters which will solidify France’s position as a global leader in AI research, development and talent.

Based on these solid foundations, we expect to see new AI ventures in France in healthcare, industrial automation and creative industries, such as video games.

AI and IP

AI patents are growing rapidly. According to EPO data, the number of international patent families for core AI rose from 41 in 2008 to 1109 in 2018. However, patenting AI and ML inventions can be challenging, especially in Europe.

AI, as a subset of computer science, falls under the category of computer-implemented inventions (CIIs). These inventions, involving computers or programmable devices, are subject to specific patentability criteria. While software itself is generally excluded from patent protection, inventions incorporating software, including AI, can be patented if they demonstrate a technical character.

The EPO has developed clear guidelines for assessing the patentability of CIIs, including those involving AI. To be patentable, an AI invention must not only meet the criteria of novelty and inventive step, but also provide a technical solution to a technical problem. This technical contribution can be achieved through the application of AI to solve a specific problem in a technical field or through a technically advanced implementation of AI algorithms.

The EPO’s examining divisions and its boards of appeal use the two-hurdle and COMVIK approaches to assess the patentability of AI inventions, as confirmed in Decision G1/19 (Pedestrian simulation) by the Enlarged Board of Appeal. Recent versions of the EPO’s guidelines for examination also include these criteria.

This framework allows for patent protection in sectors utilising AI, such as medical devices, automotive and aerospace, putting such ventures at an advantage when attracting funding. However, in AI, it appears that traditional IP, such as patents and trade marks, is being combined with the development of proprietary algorithms and with the management of a large amount of proprietary data.

Established industrial firms or luxury brands that wish to enhance their operations by adopting AI face similar challenges with their IP. To remain competitive, they will rely on finding the right combination to manage their algorithms, data and patented innovations.

AI will transform the professional practice of IP as well. For law firms, new AI software such as Harvey, is currently being deployed at scale in large law firms in New York, London and Paris.

This kind of software acts as an assistant for lawyers, notably for legal search or the preparation of a first draft on a legal matter. Lawyers in Paris have found that these tools are reducing the time spent revising commercial contracts, for instance, allowing them to improve their level of service.

Also, for patent practices, AI services such as ipQuants, Dolcera, Solve Intelligence or Qatent (acquired by Questel) are proposing new tools to perform tasks, such as drafting patent applications or responding to developments in a case. In the next three to five years, the daily work of IP professionals is likely to be transformed. IP managers in high-growth ventures will also gain from these improvements in productivity and efficiency.

• 'AI, its ecosystem and IP', an article by Samuel Deschamps, first appeared in Managing Intellectual Property Today, 2025 edition, published by Novaro, ISBN: 978-1-0685644-1-3. See here for further details.