Innovation Law Insights

Artificial Intelligence

EU Commission clarifies definition of "AI System": Guidelines published

On 6 February 2025, the European Commission published interpretative guidelines to clarify the definition of an “AI system” under Article 3(1) of Regulation (EU) 2024/1689 (AI Act).

The guidelines respond to the requirement of Art. 96(1)(f) of the Regulation and offer valuable guidance on whether a system can be considered AI for the purposes of the AI Act. The stated aim is to facilitate effective implementation of the first provisions of the Regulation that came into force on 2 February, including those relating to AI literacy and prohibitions on certain practices considered too risky for fundamental rights.

Constituent elements of an AI system

The guidelines clearly break down the elements (seven, individually treated) that characterise the definition of an “AI system” in Article 3(1) AI Act. It’s emphasised that these elements can manifest themselves differently in the two main phases of an AI system's life cycle: the building phase (pre-deployment) and the use phase (post-deployment). It’s not necessary that all elements are continuously present in both phases.

Here we summarise the different elements:

• Machine-based system

The term machine-based (as already clarified in Recital 12) refers to the fact that AI systems are developed with and run on machines, which includes both the hardware and software components needed to operate them.

• Autonomy

The system is designed to operate with “some degree of independence” from human intervention. In clarifying the scope of the concept expressed in Recital 12, the reference is intended to exclude systems designed to operate exclusively with human involvement and manual intervention.

• Adaptiveness

Intrinsically linked to the autonomy element, system adaptability can include its self-learning capabilities, which can manifest itself through changes in system behaviour, generating different outputs in response to the same inputs over time.

It’s made clear how, on the other hand, a system might not necessarily have to possess adaptiveness or self-learning capabilities after deployment to constitute an AI system.

• System objectives

An AI system is typically designed to achieve one or more objectives, which can be explicit (because they refer to directly encoded goals by the developer into the system) and implicit (not explicitly stated, but inferable from the system's behaviour or basic assumptions).

If the objectives of the system are internal to it, its so-called intended purpose is externally oriented and includes the context in which the system is designed to be used and how it must be operated.

• Inferring how to generate outputs using AI techniques, which can influence physical or virtual environments

The system must be able to infer how to produce an output from the input it receives. This capability is the key characteristic for distinguishing AI systems from other types of systems.

The terms “infers how to,” used in Article 3(1) and clarified in recital 12 of the Regulation, isn’t limited to a narrow understanding of the concept of inference as the ability of a system to derive an output from an input, and infer the result.

Consequently, the wording used, ie “infers how to generate outputs” should be understood as referring to the building phase, in which a system derives output through AI techniques (which include machine-learning, supervised/unsupervised learning, deep learning) enabling inference.

The outputs of an AI system can take different forms: predictions, content, recommendations or decisions that can affect a physical environment (eg the movement of a robot) or a virtual one (digital spaces, data flows).

Cases excluded from the definition

The guidelines also clarify which systems are not to be considered AI under the definition. They include:

• Basic data processing: systems developed and deployed to perform tasks based on manual inputs or rules, without any “learning, reasoning or modelling” at any stage of the system life cycle.
• Systems based on classical heuristics: systems that solve problems with predetermined logic, without changing their behaviour based on new data.
• Simple predictive systems: although technically they can be considered to be based on machine-learning approaches, all machine-based systems whose performance can be achieved through basic statistical learning rules are also excluded.

Conclusions

Although non-binding, the guidelines provide an important first tool to help providers and other stakeholders in assessing whether software falls within the definition of an "AI system" under the AI Act.

Author: Gabriele Cattaneo

 

Intellectual Property

The Glashütte case: Lack of distinctiveness of a geographical mark for virtual goods

In a recent ruling, the General Court (GC) addressed the substantial difference between virtual and physical goods and the criteria for assessing the distinctiveness of a geographical trademark intended for virtual goods.

The decision arose from an appeal against the decision issued by the Fifth Board of Appeal of the European Union Intellectual Property Office (EUIPO) in case R 773/2023-5. The appellant is a company belonging to a well-known German luxury watchmaking group based in the German city of Glashütte.

The trademark application at the centre of the dispute concerned a sign that included the name of the city. Additionally, the application referred to virtual goods, including watches, in class 9, as well as related retail services and supply services in classes 35 and 41.

This ruling is of particular interest for the conclusions reached by the GC regarding issues related to a geographical name and the lack of distinctiveness in relation to the protection of a trademark for virtual goods.

Issues with geographical names as trademark

The Board of Appeal recognised that “Glashütte” is a well-known German city, famous for its watchmaking tradition, and that the German public would immediately associate the name with the production of high-quality watches. This association is sufficient to render the term “Glashütte” devoid of distinctiveness under Article 7(1)(c) of the European Union Trade Mark Regulation (EUTMR), which excludes trademarks composed solely of signs or indications that may serve to designate the geographical origin or other characteristics of the goods or services. For this reason, the GC confirmed the lack of distinctiveness of the sign for marking watches.

Distinctiveness of a trademark for virtual goods

Regarding the assessment of distinctiveness and the virtual nature of the goods, the GC ruled that, in principle, the relevant public will perceive virtual goods and services in the same way as they perceive their corresponding physical goods and services. Each case should be evaluated on its specific facts, taking into account the nature of the virtual goods and services involved. But if the virtual goods are simply reproductions of real goods, or if the virtual goods and services emulate their functions, it’s highly likely that the public's perception of the virtual goods will be merely a projection of how they would perceive the corresponding real goods.

Building on this, the GC concluded that, since the virtual watches replicate and emulate the functions of real watches, the public's perception of the trademark for virtual goods would be influenced by the notoriety of Glashütte in the traditional watchmaking sector, just as it would be for a trademark intended for real goods. The sign, therefore, is viewed solely as promotional information about the quality and authenticity of the virtual goods and services, and as such, it was deemed devoid of distinctiveness.

Author: Tamara D’Angeli

 

Legal Design Tricks

Trick #6: Prototyping – From idea to reality!

Turn your idea into reality before investing, and refine it before implementation.

What is prototyping?

Creating a prototype is the first step towards producing a final, implementable solution.

In design thinking, prototyping means developing tangible models that simulate a product or service to test its value. In legal design, it means experimenting with new document formats, contracts, or legal processes to assess usability before implementation.

Why is prototyping essential?

• It helps visualize and test ideas.
• It turns abstract concepts into tangible solutions.
• It prevents wasted time and resources on ineffective solutions.

A flawed prototype today is better than a failed solution tomorrow!

What are the key principles of effective prototyping?

The golden rules:

• Fail fast – The sooner you fail, the faster you improve.
• Iterate quickly – Rapid iteration leads to better solutions.
• Minimum Viable Product (MVP) – Start with an essential version, then refine.
• Keep it simple, stupid (KISS) – The simpler the prototype, the more effective the outcome.

The goal? Identify what works (and what doesn’t) as early as possible!

How do you prototype?

Legal Design uses various prototyping techniques inspired by the design world. Here are a few:

• Wireframes – Sketch simplified layouts to test your information architecture.
• Storyboards – Develop visual sequences to simulate the user experience of a document or service.
• Interactive mockups – Create digital versions of your project to test navigation and usability.
• Role play – Simulate real-life interactions to observe how users engage with the solution.

A good prototype is flexible, adaptable, and always open to improvement!

Let’s look at an example!

Remember the client we helped generate ideas for last month? Now they want to improve their contract templates to make them clearer and more intuitive.

1. Step 1 – The legal team creates a Minimum Viable Product: a simplified draft with icons and explanations, tested internally for instant feedback.
2. Step 2 – They apply fail fast and iterate quickly, making rapid adjustments based on user feedback.
3. Step 3 – They follow the KISS principle, removing unnecessary complexity and ensuring clarity.

The result? A contract that’s clear and effective from day one, leading to smoother negotiations!

Did you know?

Before launching a new product or service, Amazon writes a press release first. It’s called the "working backward" method!

Why? If they can’t clearly and compellingly explain the product’s value, it’s not ready for the market.

In Legal Design, this means that if your contract doesn’t explain itself, it needs a redesign! KISS and iterate quickly!

What’s next?

You have a prototype. But does it actually work? The next challenge is testing!

Stay tuned for the #7 of Legal Design Tricks!

Author: Deborah Paracchini

 

Technology Media and Telecommunication

AGCom consultation on rules for frequencies in the 24.25 – 26.5 GHz band

On 27 January 2025, AGCom published Resolution No. 21/25/CONS, which launched a public consultation on the procedures for the assignment and rules for using available frequencies in the 24.25-26.5 GHz band, aiming to promote the development of ultra-broadband wireless networks.

The consultation follows the procedure started last July with Resolution No. 258/24/CONS, concerning the procedures for the assignment and the rules for the use of available frequencies in the 24.25 – 26.5 GHz band for terrestrial electronic communications systems. The initiative was also launched in view of the expiration of current usage rights assigned for this band, some of which have been extended until 31 December 2026 (while other usage rights expired on 31 December 2024 as their expiration date wasn’t extended).

The purpose of the consultation is to gather input from interested parties on the procedures for the assignment and rules for the use of frequencies in the 24.25 – 26.5 GHz band (known as the "Lower 26 GHz Band"), to support the development of 5G networks, both mobile and Fixed Wireless Access (FWA).

The Lower 26 GHz Band is part of the 24.25 – 27.5 GHz band, which has been included among the pioneer bands at the European level for the development of 5G technology. As indicated in Annex A of the resolution, part of this frequency range is currently used for point-to-point and point-to-multipoint systems for wireless local loop (WLL) services. However, the characteristics of the Lower 26 GHz Band make it particularly suitable for providing high-capacity electronic communications services, such as those based on 5G technology.

The consultation document, as outlined in Attachment A to Resolution No. 21/25/CONS, presents the Authority's proposals for the assignment and use of frequencies in the Lower 26 GHz Band.

After an introductory section, the Authority examines the relevant regulatory and legislative context, with a particular focus on the conditions for using the Lower 26 GHz Band, the framework for assigning WLL usage rights, the technological context, development needs, and the current occupation of the band by other services.

In the next section, AGCom presents its guidelines regarding the assignment procedures and rules for using the Lower 26 GHz Band, including the draft measure that AGCom proposes to adopt to regulate the assignment and use of the frequencies included in the Lower 26 GHz Band.

The provision concerns the procedures for granting usage rights for the available frequencies in the Lower 26 GHz Band, aimed at supporting the provision of public terrestrial ultra-broadband electronic communication services, as well as the related conditions for using these frequencies.

The draft measure describes the procedures for assigning WLL usage rights for the Lower 26 GHz Band, with a specific focus on the requirements for submitting applications to participate in the process.

A specific procedure is also foreseen for cases where the frequencies aren’t assigned after the first phase of the procedure or at the end of the auction.

The draft text also outlines the conditions and general obligations for the orderly, efficient, and non-interfering use of frequencies in the Lower 26 GHz Band.

It also includes the possibility for current WLL usage rightsholders to benefit from a transitional usage period for the frequencies until 31 December 2027, to facilitate the transition from WLL systems to ultra-broadband systems.

Interested parties wishing to participate in the public consultation can submit their contributions by 13 March 2025.

Authors: Flaminia Perna, Matilde Losa

---

Innovation Law Insights is compiled by DLA Piper lawyers, coordinated by Edoardo Bardelli, Carolina Battistella, Carlotta Busani, Giorgia Carneri, Noemi Canova, Gabriele Cattaneo, Maria Rita Cormaci, Camila Crisci, Cristina Criscuoli, Tamara D’Angeli, Chiara D’Onofrio, Federico Maria Di Vizio, Nadia Feola, Laura Gastaldi, Vincenzo Giuffré, Nicola Landolfi, Giacomo Lusardi, Valentina Mazza, Lara Mastrangelo, Maria Chiara Meneghetti, Deborah Paracchini, Maria Vittoria Pessina, Marianna Riedo, Tommaso Ricci, Rebecca Rossi, Roxana Smeria, Massimiliano Tiberio, Federico Toscani,  Federico Toscani, Giulia Zappaterra.

Articles concerning Telecommunications are curated by Massimo D’Andrea, Flaminia Perna, Matilde Losa and Arianna Porretti.

For further information on the topics covered, please contact the partners Giulio Coraggio, Marco de Morpurgo, Gualtiero Dragotti, Alessandro Ferrari, Roberto Valenti, Elena Varese, Alessandro Boso Caretta, Ginevra Righini.

Learn about Prisca AI Compliance, the legal tech tool developed by DLA Piper to assess the maturity of AI systems against key regulations and technical standards here.

You can learn more about “Transfer”, the legal tech tool developed by DLA Piper to support companies in evaluating data transfers out of the EEA (TIA) here, and check out a DLA Piper publication outlining Gambling regulation here, as well as a report analyzing key legal issues arising from the metaverse qui, and a comparative guide to regulations on lootboxes here.

If you no longer wish to receive Innovation Law Insights or would like to subscribe, please email Silvia Molignani.