AI requires a rapid evolution of the legislative framework in order to maintain real and effective protection for creators, says Pascal Faure, Chief Executive Officer of INPI – French Patent and Trademark Office. Pascal was in Bucharest in December, attending the regional conference "Intellectual property in the artificial intelligence era", part of the activities of the Francophone Alliance of Intellectual Property. The event was attended by representatives of French, Romanian, Moldovan and Bulgarian institutions, as well as academics, magistrates and specialists in the field of intellectual property.
"AI is profoundly transforming the ways of creation and innovation processes and raising new legal questions, to which offices, lawyers, magistrates and researchers must provide coherent answers", says Pascal.
We talked to Pascal about the stage we are currently at in terms of the process of regulating AI in relation to intellectual property, what type of difficulties creators face and how we can reach a better understanding and delimitation of the two, in an era in which artificial intelligence is shaping the landscape from one day to the next.
What led to the organization of this conference
Several reasons explain why we organized this conference. First, it is part of the momentum driven by the Francophone Alliance for Intellectual Property (Alliance francophone de la Propriété Intellectuelle), launched in October 2024 on the sidelines of the Francophonie Summit. For over a year now, this Alliance has demonstrated its ability to bring together Francophone stakeholders in innovation and intellectual property through meetings, conferences, webinars, and initiatives carried out with partners such as the AUF (Agence Universitaire de la Francophonie). In this context, it seemed essential to keep this momentum going by organizing a regional conference in South-Eastern Europe, bringing together experts from France, Romania, Moldova, and Bulgaria. Holding the event in Bucharest has strong symbolic value: Romania is a country deeply committed within the Francophonie.
This conference is all the more important because it is built on a strong collective organization, bringing together the French Embassy and the French Institute in Romania, the National Institute of Industrial Property, the Faculty of Law of the University of Bucharest, and the Franco-Romanian Legal College of European Studies, with the support of the Francophone Alliance for Intellectual Property. Their joint mobilization demonstrates a shared desire to structure a regional space for dialogue on these issues.
Lastly, the timing is strategic. The topic of intellectual property in the age of artificial intelligence is now emerging as a priority. AI is profoundly transforming modes of creation and innovation processes, and it raises unprecedented legal questions to which IP offices, lawyers, judges, and researchers must provide coherent answers.
Specifically, we will address four main areas:
- The patentability of AI-related inventions, including key issues such as inventive step, technical criteria, and the status of the inventor.
- The protection of AI-generated creations under copyright law: the notion of originality, the extent of human contribution, and the current limits of protecting non-human works.
- Issues related to data—the fuel of AI: GDPR, the text and data mining exception, database protection, algorithmic transparency, and trade secrets.
- Emerging judicial and doctrinal trends in Europe and in the participating countries, to foster constructive dialogue among practitioners, IP offices, judges, and academics.
This conference therefore aims to secure and support innovation in a context of accelerated transformation, while strengthening regional and Francophone cooperation on intellectual property.
Where do we stand in terms of intellectual property at a time when AI seems to be gaining more and more ground
We are in a phase of transition and active observation. To date, the fundamental legislation remains unchanged: IP law is designed to protect human creations. For now, there is no “special right” or legal personality for AI.
However, the application of these rules is being put to the test. Regarding AI as an inventor, the distinction is now made between AI as an assistance tool (where the human remains the author/inventor) and creations generated exclusively by AI (not recognized at this stage). The current urgency is twofold: to adapt the interpretation of existing texts and to work towards international harmonization, because AI knows no borders.
Main problems or grey areas
The challenges arise at three levels:
- At patent office level: We are seeing an increase in filings (around 6% in France concern patent applications involving the use of AI in the invention). Another challenge is the growing volume of prior art (AI-generated publications), which complicates prior-art searches. Training examiners in these new technologies is also important.
- A growing share of applications and publications drafted with AI assistance is unclear and sometimes poorly understood by their authors (for example, an individual asking ChatGPT to draft their invention idea without mastering the underlying technology).
- Ownership of rights: The central question remains: Who is the inventor? The current consensus (e.g., DABUS case law) refuses to designate an AI as inventor, but this raises issues about remuneration and ownership of automatically generated results.
- Training data: This is the major grey area. Transparency regarding the data used to train models (copyright, trade secrets) and the sufficiency of disclosure in patents are key topics.
To what extent can AI be regulated
Regulation is necessary but complex for two main reasons:
- Time lag: Technology evolves exponentially (“engineers’ time”), whereas the legislative process is by nature slower and more deliberate (“the law’s time”).
- Territoriality vs. universality: IP is territorial (a patent is valid in one country), while AI is digital and global. A model trained in the United States (under Fair Use) can be used in Europe, but its owner must comply with EU rules (where transparency requirements are stronger), creating jurisdictional conflicts and competitive distortions.
Which areas are most affected by the rise of AI
All fields are affected, but two pillars of IP are on the front line:
- Copyright: On the one hand, the massive use of protected works for text and data mining without explicit consent; on the other, the protection of generated works.
- Personality rights and image rights: With deepfakes (audio and video), artists’ image and voice are cloned, which goes beyond IP and touches on fundamental personal rights.
What kinds of problems have already appeared
Conflicts are multiplying and fall into two categories:
- Training: Major collective actions (class actions), particularly in the United States and Europe, pit creators (writers, visual artists) against AI providers (OpenAI, Stability AI, etc.). The main complaint is the unauthorized and unpaid use of their works to train models.
- Use: The proliferation of content imitating living artists’ style or using their voice (e.g., fake songs by famous artists), as well as misinformation (fake news) issues that harm reputation.
Where are we at the European level regarding IP protection
Europe confirms its position as a global pioneer with the AI Act (Regulation on AI), while remaining pragmatic so as not to hinder innovation. The foundation of the AI Act (transparency and copyright) is the crucial point that remains unchanged: Europe imposes strict rules on providers of general-purpose AI (such as OpenAI or Mistral). They are required to comply with European copyright law and to publish a sufficiently detailed summary of the data used to train their models. This measure is fundamental to enable rights holders to check whether their works have been used.
In November 2025, the European Commission proposed a package of simplification measures (“Digital Omnibus”) to reduce the administrative burden on businesses: this involves postponing compliance deadlines for “high-risk” systems (e.g., to the end of 2027), and certain reporting or training (“AI literacy”) obligations are relaxed, especially for SMEs.
The transparency obligations regarding training data and respect for intellectual property are not being called into question.
Alongside legislation, harmonization will also come through judges. Countries such as Germany (LAION and GEMA decisions) or France are beginning to define, through their courts, the limits of text and data mining.
New disciplines that address the AI–IP relationship
Specialized programs (such as at CEIPI in Strasbourg) already include “AI & IP” modules.
The future goal is to develop dual expertise. Engineers must be made aware of intellectual property (ethics, infringement), and lawyers must understand how AI works technically. This cultural shift is necessary to prepare tomorrow’s professionals.
Your message to those who currently feel threatened by AI’s interference
IP law continues to protect human creations, considering AI as a tool. However, given the unprecedented capabilities of these models, AI requires a rapid evolution of the legislative framework in order to maintain real and effective protection for creators.
The added value of content creators (videos, presentations, literary works, etc.) lies in inspiration, the creative way of delivering the right idea to the right person, and in the choices and trade-offs they make.
AI mainly helps optimize the work needed to produce content, while the creative aspect of the work remains always linked to the creators themselves.
