Sound Marks, Deepfakes, and the Identity Gap: What Meloni's EUIPO Voice Filing Means for Brand Owners

The filing that changed the conversation

On 5 October 2026, Italian Prime Minister Giorgia Meloni submitted a formal sound-mark application to the European Union Intellectual Property Office. The filing centres on a four-second audio clip in which Meloni twice repeats the phrase "Io sono Giorgia Meloni" - a line that first became politically iconic at a 2019 rally and later gave its shortened form to her 2021 autobiography. Her office confirmed the application is a direct response to AI-generated deepfakes; it was filed through Rome firm Mpmlegal in coordination with Praxi Intellectual Property S.p.A. As of 9 October, the mark remains under formal examination by EUIPO examiners.

Meloni is, according to the Financial Times, the first major politician known to seek trademark protection for her voice. She joins a short list of entertainers, including Taylor Swift and Matthew McConaughey, who have pursued similar routes. The move follows a pending Italian civil case in which she sought €100,000 in damages against a man accused of creating deepfake pornographic images using her likeness.

What classes are actually covered?

The application spans three service categories: downloadable multimedia content, cultural activities, and the organisation of political events, mapping roughly onto Nice Classes 9, 41 and 45. That class selection is telling. This is not a broad commercial sweep. It is a targeted defensive perimeter around the specific contexts in which her voice carries legal and electoral weight.

For commercial brand owners, that precision is instructive. Sound marks are notoriously difficult to register: EUIPO examiners require demonstrable distinctiveness, and a voice clip that essentially says "I am [name]" runs a real risk of being found descriptive. Italian IP commentators have already noted that even a granted mark would not automatically block AI voice synthesis. It would create additional legal obstacles, not an impenetrable wall.

The agentic commerce problem underneath

Meloni's filing arrives as a separate, structurally important threat matures in the commercial sphere. World Trademark Review this week published analysis warning that AI shopping agents are actively accelerating counterfeit sales. The core problem: as AI agents take over purchasing decisions on behalf of consumers, brand owners must make authenticity machine-readable or risk being invisible, or worse, spoofed.

This is not a theoretical threat. AI detection agents already scan millions of marketplace listings continuously, comparing them against genuine SKU-level product data and catching trademark-evasive entries that conventional keyword searches miss entirely. The enforcement gap is widening because the counterfeiters have also upgraded: generative AI now produces convincing fake storefronts, multi-language copy, fabricated reviews and product imagery in minutes. Many such sites operate briefly, collect payments, and relaunch under a new domain before any manual monitoring cycle can close.

What the filing data signals

The timing is pointed. According to AIBD analysis of IPO trademark data, UK Class 35 filings, covering advertising, business management and retail services (the class most associated with brand positioning and commercial identity), stand at just 1,208 in Q4 2026, a sharp drop of 89.3% against the prior period. (Source: AIBD analysis of IPO (TMD) data, as of October 2026.) Taken in isolation, that figure might suggest brand owners are pulling back from formal registration. The likelier reading is more nuanced: the fee shock from UKIPO's April 2026 increases, which raised trademark fees by an average of 25% in the first such rise since 1998, created a filing surge ahead of April that has left Q4 comparatives looking dramatically depressed.

The practical danger is that, during any period of lower filing activity, the register becomes patchier. Patchy registers are precisely the environment in which bad actors file opportunistically or infringe with reduced risk of detection.

The limits of the trademark weapon

Meloni's case shows both the utility and the ceiling of trademark law as an anti-deepfake instrument. Registration establishes priority, gives grounds for takedown requests on platforms, and creates a legal basis for civil action. It does not make deepfakes impossible. EUIPO examiners will still consider whether "Io sono Giorgia Meloni" is sufficiently distinctive as a sound mark. They may conclude the phrase is too descriptive of the speaker herself to function as a commercial identifier.

For private brands, the lesson is sharper. A corporate voice, whether a brand's distinctive audio logo, sonic identity, or a spokesperson's recorded delivery, sits in even murkier territory. Such assets live primarily in Classes 9 and 41, yet many brand owners have never registered them as non-traditional marks. If generative AI can now clone those assets at scale, the enforcement gap is not one the courts will fill quickly.

The question is not whether to file. It is whether your existing portfolio covers the non-traditional identifiers, sound, motion, trade dress, that AI is most capable of replicating. For most brand owners, the answer will be uncomfortable.

What to do this week

Audit your non-traditional marks. If your brand has a sonic identity, a distinctive voice-over, or a signature audio element that appears in advertising or customer-facing content, check whether it is registered and in which classes. Engage a qualified trademark attorney or consult TrademarkDashboard to assess coverage gaps before the next filing window.