16 June 2025

Newsletter June 2025

✨IA & INTELLECTUAL PROPERTY : TOWARDS A RECOGNITION OF THE VALUE OF CULTURAL CONTENT✨

First case on IA and copyright in front of the Court of Justice of the EU

The Court of Justice of EU (ECJ) has just received its very first preliminary ruling on IA and copyright.

The reference, registered under number C-250/25, comes from Hungary and involves a company and Google. At stake : the neighbouring rights of press publishers (Article 15 of the DSM Directive), exceptions for text and data mining (TDM – Article 4 of the same Directive), and more broadly, the way in which protected content may (or may not) be displayed by chatbots.

In particular, the Hungarian judge is asking whether the fact that a chatbot reproduces, in its responses, content that is (almost) identical to that published on a newspaper’s website constitutes an infringement :

the right of reproduction ;
the right to make content available to the public.

This is the case even if the chatbot’s response is the result of a probabilistic process: it “guesses” word after word what it should say, based on the models it has learned.

Another key question concerns the training of AI models. Should this be seen as an act of reproduction? If so, does the TDM exception provided for in the Directive apply to this massive training?

A quick reminder: Article 15 of the DSM Directive grants press publishers a neighbouring right, enabling them to control the reproduction and online availability of their publications by services such as Google. This right is directly inspired by Articles 2 and 3(2) of the InfoSoc Directive.

Article 4 requires Member States to introduce a system of ‘neighbouring rights’. But beware: this exception only covers the right of reproduction and extraction, not the right of communication to the public. And it can be set aside if the right holders expressly object (opt-out).

This case brings us back to three hotly debated issues in European doctrine:

1️⃣ Do chatbots infringe copyright?

If the content displayed is protected, and if the AI has not created the text completely independently, there is a good chance that the chatbot infringes the rights of reproduction and making available.

And it doesn’t matter if it’s a word-for-word “prediction”: if the end result is a reproduction, the legal qualification should not be altered.

2️⃣ Is AI training an act of reproduction?

The answer, already suggested by European case law, is yes. This is also the reason why the legislator has provided for a specific exception for TDM. The reproduction right is interpreted broadly, which includes technical acts linked to the massive analysis of protected data.

3️⃣Does TDM cover AI training?

Here again, even if some doubted it, the answer now seems clear. AI was one of the concerns when the DSM directive was adopted in 2019. And the explicit link between TDM and AI in the recent AI Act only confirms this focus.

But an important clarification is in order: TDM covers only part of AI training. To date, no general exception explicitly covers AI training as a whole. Everything is still based on a subtle interplay of exceptions, conditions (lawful access, specific purpose, etc.) and tests (three-step test).

Find out more 👉 https://curia.europa.eu/juris/showPdf.jsf?text=&docid=300681&pageIndex=0&doclang=EN&mode=req&dir=&occ=first&part=1&cid=5279466

This first case opens a new chapter, in which copyright and artificial intelligence will have to learn to live together… under the watchful eye of European judges.

An European Code of good practices on AI to the detriment of copyright?

On 14 May, the French Senate Committee for European Affairs raised an issue.

While the IA Act was adopted with the ambition of guaranteeing the transparency and fairness of AI systems, the Senate denounced a worrying weakening of its fundamental principles. At stake: a first version of the code of good practices that appears to give disproportionate weight to the interests of tech giants… to the detriment of European creators.

Generative AIs are not neutral. They feed on protected works, press content and artistic and literary creations. And this massive exploitation takes place without authorisation or remuneration, jeopardising the already fragile economic models of the cultural industries.

In a political opinion adopted unanimously, the senators call on the European Commission to review its draft Code of Practice on artificial intelligence and set out clear requirements:

➡️ No work should be used without the express permission of its authors;

➡️ The principle of transparency, a pillar of the IA Act, should be maintained and strengthened;

➡️ All stakeholders – creators, publishers, producers, journalists – should be represented at the negotiating table.

This debate echoes the report by the Conseil Supérieur de la Propriété Littéraire et Artistique (CSPLA)….

Transparency of generative AI and remuneration for cultural content:

On 23 June, the French Higher Council of Literary and Artistic Property (CSPLA) will be presenting its draft report on artificial intelligence and remuneration for cultural content.

Today, generative AIs are massively consuming protected works available online for training purposes, without authorization, without transparency… and above all without remuneration for creators, publishers and producers.

Faced with this situation, the CSPLA report proposes a number of concrete ways of correcting the situation:

1️⃣ Transparency of sources: imposed by the IA Act, this would finally enable rights holders :

check their opt-out
and to draw economic consequences, via possible remuneration.

2️⃣ Recognition of the economic value of cultural content in AI models:

Proportional and/or flat-rate remuneration,
Creation of a structured marketplace to contractualise access to works,
Structuring databases in conjunction with rights holders.

This ambitious project aims to reconcile technological innovation and the fundamental rights of creators, while laying the foundations for a more balanced and sustainable market in cultural data.

✨ PERSONAL DATA ✨

➡️SOLOCAL fined: CNIL raises its voice on commercial prospecting and consent!

On 15 May 2025, the CNIL fined SOLOCAL €900,000, with a penalty of €10,000 per day until the company complies. Why was this? Because consent must be genuine and can be proven.

SOLOCAL used data from online forms (competitions, product tests, etc.) supplied by data brokers and website publishers. However, the forms were designed in such a way as to strongly encourage the user to accept them, by playing on the ergonomics, colours, size of buttons, etc. And above all, there was no tangible proof of acceptance.

Above all, the CNIL was unable to obtain any tangible proof of consent.

The CNIL has identified two major shortcomings:

1️⃣ Lack of a valid legal basis for canvassing (article L.34-5 of the French Post and Electronic Communications Code: consent must be free, informed and specific. In this case, it was… suggestive.

2️⃣ Lack of proof of consent (article 7 GDPR): when it comes to personal data, the data controller must be able to prove, not just hope.

Even after identifying the breaches, SOLOCAL continued to use the data for 17 months. Enough to make any supervisory authority jump for joy…

🔎 Something to remember for all companies that use “ready-made” prospecting files:

You are responsible for the data you process, even if you did not collect it yourself.
Check the forms, demand proof of consent, suspend any questionable processing.

📣 FIRSH expert voice: The GDPR is not a formality: it is a legal and strategic requirement. FIRSH can help you secure your prospecting campaigns, both upstream and downstream, so that compliance does not become an issue… once the fine has been imposed.

To read the full decision, click here 👉 Délibération SAN-2025-001 du 15 mai 2025 – Légifrance

➡️META trains on European data:

Since 27 May, META has been using the data of European users of its social networks (Facebook and Instagram) to train its artificial intelligence systems.

META has specified that the data of Meta AI users on WhatsApp is not affected.

What types of data are used?

Data contained in the public publications of adult users (texts, photos, comments, etc.)
Data from users’ interactions, whatever their age, with AI services, for example information entered into its conversational agent (AI chatbot).

Following discussions with the European authorities and the Irish Data Protection Authority (DPC), Meta has strengthened its filtering measures to reduce the risk of personal data being stored by AI models during the training phase.

✨INTELLECTUAL PROPERTY✨

➡️Trademarks : T’Choupi is a reputed trade mark within the meaning of Article 8(5) EUTMR

In this opposition case, the EUIPO recognises the reputed character of the trademark T’CHOUPI for printed matter (class 16), which enables it to oppose an application for the trade mark CHOUPI filed in classes 5 (pharmaceuticals, hygiene), 31 (pet food), and 35 (sales services).

Even though the goods concerned are not identical, the EUIPO considers that the average consumer – a parent, potentially a customer of T’choupi books and pet or hygiene articles – could make an economic link between the two trademarks. This is sufficient to characterise a risk of unjustified exploitation of T’Choupi’s reputation.

What this means:

The products do not have to be identical: there just has to be a sufficient link in the public mind.
Reputation gives a reputed trademark an ‘extended shield’, even in the face of trademarks that do not exactly imitate but evoke.
The affective/familiar dimension of the brand plays a crucial role in analyzing the mental link.

To read the full decision, click here 👉 https://euipo.europa.eu/copla/trademark/data/018845686/download/CLW/OPP/2025/EN/20250522_003201108.doc?app=caselaw&casenum=003201108&trTypeDoc=NA

➡️Trademarks : Infringement action and limitation in consequence of acquiescence:

In a decision dated 19 March 2025, the Paris Court of Appeal reiterated the scope of the limitation by acquiescence provided for in article L. 716-4-5 of the French Intellectual Property Code (FIPC).

In this case, the owner of an earlier trademark had his infringement action declared inadmissible for failing to take action for more than five years, even though he was aware of the continuous and public use of a competing later trademark.

The Court held that all the conditions laid down by the ECJ had been met: filing in good faith, actual use, registration in the State concerned, and knowledge of that use by the owner of the earlier trade mark.

This decision confirms that, when it comes to trade marks, prolonged inaction can be costly.

📣 FIRSH expert voice: defending your trademarks is an investment. FIRSH recommends that you set up a trademark protection watch so that you can take action as early as possible, and therefore at a lower cost, either during the opposition period of the trademark registration.

To read the full decision, click here 👉 CA de Paris, 19 mars 2025, n°24/03156

✨ECOSYSTEM FIRSH✨

With its ‘Choose Europe to Start and Scale’ plan, the European Commission is setting out a clear ambition: to simplify procedures, provide better funding for innovation, and create the conditions for the emergence of European champions.

A clearer framework for innovation and growth in Europe: Valuing intellectual property, supporting scale-ups, simplifying administration: this new European direction resonates with our practice. At Firsh, we’re ready to support projects that want to make a real impact, both in Europe and internationally.

Recommandation: Camille Vever’s podcast ‘Les voix de l’audace’.

In 2021, Camille and Damien Vever, the 7th generation of the family, are reawakening the House that has been dormant since 1982 with the ambition of reinventing its value proposition and adapting it to the environmental and societal challenges of the 21st century. Camille invites people who have not been afraid to be daring in their field – to innovate, to break with codes and to show a new path to excellence. Favourite FIRSH podcast! Come and listen to the podcast! For each episode, La Maison is offering 5 people the chance to attend the recording in the Vever showroom. All you have to do is send an email to contact@vever.com with the subject line “Voix de l’audace”, giving your full name and telephone number.

✨ NEWS ‘ FIRSH ✨

➡️ FIRSH has recently advised its clients on the following matters:

IP audit as part of the acquisition of a company specializing in software development
Negotiation and drafting of an R&D collaboration agreement between two companies
Drafting of a letter of formal notice in connection with unauthorized use of trademarks (counterfeiting and unfair competition)
RGPD audit of a company specializing in raw materials traceability software
Pleading before the Toulouse Court of Appeal about the questionable perimeter of a withdrawal of products from the e-market under penalty

➡️ As part of its innovation laboratory, FIRSH has published its first White Paper! This marks the official launch of FIRSH LAB, which is publishing its first study on a high-stakes social issue: deepfakes (fake images, videos, or audio generated by artificial intelligence). Addressing one of the major issues facing society in the future, this White Paper required more than 6 months of in-depth work and an in-depth study of the technical, sociological, economic and legal aspects of deepfakes, involving numerous French and international experts in artificial intelligence. The legal thinking that emerges is based on a detailed analysis of legislation, legal doctrine and court rulings. The White Paper provides legal and practical recommendations for public authorities, businesses and anyone interested in the subject.

Please contact us at the following e-mail address to receive a copy of the White Paper 👉 contact@FIRSH.LAW

📢 To follow us on LinkedIn and receive our newsletter, click here : https://www.linkedin.com/company/firshlaw/ .

📢 There is no direct collection of your personal data and therefore no emailing from Firsh!