Giorgia Meloni's Voice Trademark: Can It Stop AI Deepfakes?
Four Seconds, One Sentence : Why Giorgia Meloni Is Trying to Trademark Her Own Voice
Italy's Prime Minister has requested the EU to recognize four seconds of her speech as a trade mark, and it's not just because she's so limited with the law. On 5 October 2026, Giorgia Meloni applied for the sound mark with the European Union Intellectual Property Office (EUIPO). In effect, it's an attempt to put a lawful fence around her own voice when anybody with a laptop can clone it.
The filing is minimal, but the questions raised are huge. Can a voice be legally owned or exclusively controlled? But can trade mark law, designed to safeguard brands in the market, fill the void left by privacy law, personality law and copyright law when it comes to protecting AI generated sounds? So, what is a registered sound mark going to prevent its holder from doing?
In this article, the author details the elements of the EUIPO record, how sound marks operate in the EU, why public figures are now using them and the strengths and weaknesses of the approach. The gist: A voice mark is a helpful tool, but it isn't a protection from deepfakes. It's a worthwhile read for those working with voices, brands or AI.
Note on sources: the information contained in this register is derived from the EUIPO file (EUTM 019431219) on screenshots, which were posted on and the rest of the information is from the press coverage cited at the end. This is an application that is still being examined and not a right granted.
The actual complaint submitted by Meloni
The EUIPO register refers to the application as a "Trade mark without text", indicating that the mark is not a word or logo. The table below shows the important information contained in the file.
What is displayed on record
Sound (nature: individual)
Language of filing / second language
Giorgia Meloni, a natural person, with an address in Rome
Praxi Intellectual Property S.p.A., Civitanova Marche
Four seconds, downloadable from the register
Application under examination
The application is flagged as meeting Fast Track conditions
Press reports, citing the EUIPO records and her office, say the recording is Meloni saying "Io sono Giorgia Meloni" ("I am Giorgia Meloni") twice in Italian. It is a reference to the title of her 2021 autobiography Io sono Giorgia, which originated from a quote during a 2019 rally, and was later turned into a dance track.
There are three points the filing to note
The applicant is not a party but Meloni personally, or any government body. This is important because a trade mark must be used or be used by the owner of it or by his consent and in the course of trade. By filing as an individual, she is making it clear that she is using her voice not as a logo for her institution, but as a brand for herself, like a celebrity.
Secondly, the goods and services are carefully selected. Class 9 is for downloadable and recorded Media content, Multimedia files and Electronic Publications. Class 41 includes cultural activities, conferences, seminars, publishing and related services. Classes 45 are listed in the record with items like organising political meetings and legal research within the public policy field. In other words, the claim is grounded in what she already knows as a media, publishing, events and public-affairs name and voice, which already has commercial and political value.
Third, the application has a flag of "Fast Track". Fast Track at the EUIPO is available for applications which are based on terms of the office's pre-established database of goods and services. It may expedite the examination of the application as it was filed but does not affect the legal tests that the mark must meet.
Within the EU, how sound marks work
Sound marks are not out-of-the-way. With the EU trade mark reform, which came into effect in October 2017, there is no longer a requirement to represent a mark "graphically", so a sound mark can now be filed in the form of an audio file (e.g. MP3) or in the form of musical notation. A sound mark is a mark which contains only a sound or a combination of sounds and is registered under article 3(3)(g) of the Implementing Regulation. Do not submit a sonogram representation by itself and if it is included with an audio file, it is removed.
It made a difference to people like Meloni. Prior to 2017, one could only describe a sound in a manner that the register would be able to display, more suitable for a jingle written using notation rather than a spoken sentence in a specific human voice. Pitch, timbre, rhythm and accent can be captured in an audio file, but not in a stave. It also implies that the protection is linked to the specific recording made. Another sentence (or the same sentence in another voice), is another sign.
In every other respect a sound mark must meet the same requirements as any other EU trade mark. It has to be:
Clear and unambiguous, enabling authorities and the public to understand what's being protected. This is fulfilled by an audio file.
Distinctive (distinguished from sound) - in that it is perceived by the consumer as a sign of the commercial origin, and not solely as sound.
No absolute prohibition such as being descriptive, functional or against public policy.
Sound is most likely to be distinctive. In the Ardagh case, where the EU General Court ruled on the sound of a drinks can being opened followed by a fizz, it ruled that the sound was similar to a product characteristic rather than a badge of origin. There are similar fates for applications that make short engine noises and for generic notification sounds. The lesson: A sound which a person hears as a characteristic, not a trademark, labors.
Application is published for opposition and, if there are no objections, is registered if accepted by EUIPO. A registered EU trade mark protects against the infringement of the 27 Member States of the European Union, its duration is 10 years, and it can be renewed indefinitely provided that the trade mark is used in a genuine manner. The review of this specific app is still in progress and there is no announcement.
Why a voice, why now?
Voice cloning has progressed from a research demo to commodity. Just a couple seconds of clean audio can suffice for some tools to make a convincing copy, and a politician is a sweet target: hours of public recordings, a cadence easily recognizable, an audience who easily believe or something outrageous. Politically, Meloni has publicly discredited AI-generated lies and implored citizens to verify the content before spreading falsehoods, and elections for her office are scheduled to take place approximately a year from now in Italy. The filing has been tied to those deep fake concerns by the press.
She is by no means the first Italian to attempt this path. The singer Giusy Ferreri reportedly registered the sound mark "Sono Giusy Ferreri" in May 2026, which is the first European artist to do so, and the voice actor Luca Ward reportedly did the same. In every instance the sound is a self-referential and short sentence, "I am [name]". It's no coincidence.
The examiner of a trade mark will not accept the registration of "a voice" in the abstract. One can't register Meloni's intonation. A specific recording may be filed. A sentence that announces the speaker will serve three purposes:
It makes the sound like it's a signature, enhancing the distinctiveness.
It links the recording to a brand name which is already well known, which will aid in origin.
Provides a rights holder with a tangible standard to use if a clone utters the identical words in a similar voice.
A signaling effect also occurs. Publicly sending these platforms, publishers and potential imitators a message: This voice is ours. The filing also establishes a precedent that is likely to draw press, even if the mark is never registered, and that will change the norm: voices will be assets that people will fight for.
Last but not least, the transfer is a hole. The EU copyright framework and most national privacy frameworks do not explicitly provide any right for the person to claim a right of “owner” over their own voice. Recorded performances/works are protected by copyright, not a person’s voice itself. Trade mark law is accessible, known and easy to administer – lawyers grab it.
What a trademark is and what it can't do against an AI clone. What a trademark can and cannot do against an AI clone
But the most crucial part of the story is the least showy - no one can be prevented from making a clone of a voice with a registered sound mark. Reuters, quoting Italian media, and other outlets reported just that. A trade mark does not entitle the trademarks owner to generally ban the making of imitations rather than the use of the sign in trade.
Now when does it help? According to Article 9 of the EU Trade Mark Regulation, the proprietor may prohibit the use by third parties of a trade mark without its consent of an identical or similar sign in the course of trade for identical or similar goods or services if there is a likelihood of confusion. There's also enhanced protection for marks that have a reputation, which can be applied to other goods that take unfair advantage of, or detract from, the reputation or distinctive character of the mark.
The scenarios for Meloni are as follows
A fake audiobook or podcast with a fake voice of "I am Giorgia Meloni" and commercially sold or distributed. This is near the registered classes 9 and 41 and a confusion argument seems reasonable.
A fake video (on social media) that is used for political purposes, it does not sell anything, has nothing commercial in mind. Trade mark law is not appropriate; it cannot be said to be used in the course of trade, and the audience is not likely to see the clip as evidence of the origin of the product.
That third scenario is the one that democracies fear, and one where trade mark does least. A voice mark has the strongest impact where there is commercial exploitation, and the weakest impact where there is disinformation, which is a political/criminal-law issue, not a brand issue.
A registration still is useful for practical purposes. It facilitates easy identification, retrieval, and citation of relevant information in takedown notice. It can be adapted to requests to platforms and marketplaces with existing complaint processes for trademarks. Can be recorded with customs/enforcement bodies for physical products with the sound, like electronic devices. And it can make a bad guy's life more difficult because it provides a court with a registered sign to compare.
The application must overcome many hurdles in order to become a right and some will be relevant even post-registration.
Distinctiveness. The examiner will ask if the average consumer would perceive this sound as a cue to the source for the goods/services identified. When a name is spoken in a sentence, it will be readily understandable as plain speech. The counterargument is that the speaker is one of the best-known voices in European politics and the sound is a strong association to one source. This will likely be the big issue. Evidence of acquired distinctiveness (audience recognition) can be used but only for the appropriate audience and goods.
Names take the center stage in establishing centers : Names of living people are normally registrable as trademarks, but in doing so they are also being protected by word marks and personality rights, which Meloni can pursue. Voice is an added value of the sound mark, but voice can only be protected if a sample is filed. A sequence of words that is different from the one the speaker says, or a different cadence, may be off the registered sign.
The perception of similarity is the standard for judging similarity. The test for infringement is whether the clone is identical or similar to the registered sound as heard by a listener. Although AI- generated voices can capture the general characteristics of an individual’s voice, their ability to replicate finer nuances remain inconsistent. The courts will require expert audio comparison and there is sparse case law as to how similar the synthetic voice must be to count.
Trade as such. As mentioned, non-commercial political deepfakes will typically not infringe upon anyone's trademark. Freedom of expression and parody/satire is also available to a defendant as a defence, a defence that is balanced against property rights by the EU courts.
Questions of public office, genuine use. A registered EU trade mark which is not used for five years is at risk of being cancelled. If the sitting head of government does not have a plan for use, he will have a thin right, if he does not trade in audiobooks or events. The officeholder may also be subject to political and ethical monitoring regarding the exclusive use of his voice in public discourse, and objections to his conduct on public interest issues are always possible.
Scope will be constrained to classes filed. The following list shows the goods and services covered by protection. The more difficult to attack a deepfake will be if it is sold as a ringtone or a game and doesn't get covered or something similar to what is registered. Expanding the list means that there will be objections from the office and from third parties.
None of these points is the reason to file a useless file. They outline why expert counsel see voice marks as being only one step in a comprehensive strategy, rather than a defence in itself.
What else is in the way of Meloni getting a voice clone?
It's only part of her job profile and her attorneys will recognize it. There are a number of other routes available, but each has their own shortcomings.
The rights of personality and image rights. A person's name and likeness has always been a protected feature of personality in Italy against the use of their likeness for unauthorised commercial purposes, and it is equally argued that a recognizable voice is a personality feature and is under the same reasoning. Remedies are typically of the after-the-fact variety.
Data protection. The voice processed in order to identify a person may be considered as biometric data under the GDPR. It gives her a case for claiming they should be removed, but it's difficult to do so against anonymous uploaders or platforms not in the EU.
Criminal law. The harmful spread of AI-generated content is closer to the political deepfake scenario than any trade mark claim, and Italy has been taking steps to criminalize such content.
EU AI laws and platform guidelines - Timing and enforcement of the AI Act's obligations for transparency of synthetic content and platform regulations prohibiting impersonation can help identify or remove fakes.
The "real world" application is layering. The mark covers commercial use, the others cover the balance, and a registration can provide the paperwork that makes a platform compliant credible. The chances of a person obtaining a swift result from the impersonation report on a platform is significantly greater than from a court, which is where a registered sign earns its keep.
What to look for as the application continues
It's an early file and we'll see over the few months if the strategy will be a success or a failure. Five checkpoints matter:
Examiner's opinion of distinctiveness. The first hurdle may be an objection based on the absolute grounds for refusal; then it will have to decide whether the phrase "Io sono Giorgia Meloni" is a signature of origin or a simple statement.
How the list of goods/services perform . The territory is defined by the classes 9, 41 and 45. Any restriction on the office's demands will reveal the extent to which this mark can really serve her commercial and political activity.
The opposition window. When the mark is published, there is a limited time period for opposition by third parties. If anyone (whether it be a competitor in politics or some media outlet) challenges it, it will be telling.
An actual usage plan. Unused registered marks are susceptible. Monitoring for use of Meloni (or a licensed user) on registered fields.
Copycats. Before her in the queue is Ferreri and there are others watching. If there are more politicians that file, EUIPO's reasoning in this file could serve as a template.
The four second clip from Meloni is intended to be a statement of intent and a first step that will be followed by the implementation of the rest of the energy efficiency measures. It tells the market that her voice is "mine" and, if the application is successful, it gives her lawyers a registered sign to point to, and it places a real-life test in front of the EUIPO: can a voice be a trade mark if it's fixed in a brief recording?
The truth is "sometimes, for a limited amount of time". A voice mark can assist in preventing commercial misuse like fake audiobooks, spoofed endorsements and fake media. It can't, by itself, prevent a political deepfake, and never was intended to. What people actually need for that is the law and the right to be a copy or voice of that law, plus quick enforcement mechanisms and means of proving where the law came from is still being constructed.
Now, watch what happens to the following three points: objecting on distinctiveness; seeing if anyone objects to it after it is published; seeing if other politicians and performers do the same. Whether the mark gets to move forward, it will be a first look at Europe utilizing old IP techniques on the new technology. It will otherwise be equally informative insofar as it will demonstrate the boundaries of trade mark law, and where a new form of right must start.
Italy PM Meloni seeks EU trademark for her voice amid AI deepfake risks (Reuters, via The Star)
Giorgia Meloni files to trademark her voice against AI deepfakes (The Web)
Meloni files 4-second recording of herself as an EU sound trademark (PPC Land)
What is EUIPO, the European office where Meloni has registered her voice (Il Foglio)
Italy's Meloni follows pop stars in seeking to trademark her voice (AP, via NY1)
Hitting the right note: sound trademarks at the EUIPO (World IP Review)
Sound trademarks: first European ruling on a sound mark in an audio file (ABG IP)
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
