Summary

• Giorgia Meloni has filed an EU trademark application containing a four second recording of her voice.
• The application is under examination and covers categories including downloadable media, cultural activities and political events.
• Trademark protection could provide another legal route against some commercial misuse, but it would not create a general ban on synthetic imitation.

Italian prime minister Giorgia Meloni has applied to register her own voice as an EU trademark, turning existing intellectual property law towards a problem created by increasingly convincing AI impersonation.

The application filed with the European Union Intellectual Property Office contains a four second recording of Meloni saying “Io sono Giorgia Meloni”, the phrase associated with a widely recognised speech she gave in 2019. The filing is currently under examination rather than already registered.

According to confirmation from Meloni’s office, the application is intended to strengthen protection against unauthorised voice cloning. It covers several service categories, including downloadable multimedia content, cultural activities and the organisation of political events.

The move does not create a general ownership right over every use of Meloni’s voice, nor would trademark law automatically prevent somebody from producing a synthetic imitation. It instead tests whether a distinctive recorded voice can become another protected sign whose unauthorised commercial use may be challenged through existing law.

Voice is becoming an identifier

Generative audio technology has changed the economics of impersonation because creating a convincing synthetic voice no longer requires a professional studio or extensive recordings. Consumer and commercial tools can reproduce vocal characteristics from relatively small samples, making politicians, executives, celebrities and other people whose voices are widely available particularly straightforward targets.

The resulting risks range from parody and misinformation to fraud. A synthetic recording that appears to capture a political leader giving instructions or making a controversial statement can circulate before verification catches up, while businesses already face scams in which cloned voices are used to impersonate senior executives.

Traditional trademark law was developed mainly around commercial signs that distinguish goods and services, although those signs need not always be visual. Sound marks already exist when a sound can distinguish one undertaking’s products or services from another’s and satisfies the other requirements for registration.

Meloni’s application pushes that concept into the age of synthetic media by treating her own spoken phrase and vocal identity as something capable of serving a trademark function in specified contexts.

Whether EUIPO ultimately accepts the application, and how broad any resulting protection proves in practice, remains unresolved. Registration would also sit alongside other areas of law that may apply to impersonation, including privacy, defamation, fraud, image rights and rules around manipulated content.

A trademark protects defined uses

Establishing trademark protection over a particular recorded voice would not amount to a legal ban on copying that person’s voice in every context. Intellectual property law generally works more narrowly because a registered mark applies to designated classes of goods or services and gives its holder remedies against uses that infringe those rights.

Context consequently matters, including whether the use is commercial and whether it creates confusion or exploits the reputation associated with the mark.

Journalism, satire, political criticism and other forms of expression may therefore raise different legal considerations from an advertisement or product deliberately imitating a famous person’s voice.

AI systems complicate enforcement further because the person generating the clone, the software supplier, the platform hosting it and the account distributing it can all sit in different jurisdictions. Another potential legal claim does not automatically make the offending material easier to identify or remove.

Meloni’s use of a short and recognisable phrase may nevertheless make the application more concrete than an attempt to register an abstract claim over every possible manifestation of a person’s voice. “Io sono Giorgia Meloni” has an established association with the prime minister and has circulated in political and cultural contexts for years.

Public figures are assembling overlapping defences

Musicians, actors and other public figures have already pursued trademark, contractual and personality based approaches as generative systems make voices and appearances easier to reproduce without a conventional recording session.

Meloni has also encountered manipulated material involving her likeness and previously took legal action over deepfake pornography, placing the voice application within a broader attempt to build remedies from legal frameworks developed before modern generative AI existed.

That piecemeal response is likely to continue because no single body of law addresses every form of synthetic identity misuse. The EU AI Act introduces transparency obligations around certain generated and manipulated content, but disclosure requirements and ownership rights address different problems.

A watermark or label can tell an audience that material was generated artificially without necessarily stopping its creation. Trademark or personality rights can provide legal claims after misuse occurs without guaranteeing that audiences recognise a fake before it spreads.

Technology companies are developing provenance and detection systems in parallel, although those tools also face limitations when media is edited, rerecorded or passed through several services.

Businesses face the same identity problem

Voice is increasingly used inside commercial systems as executives appear in investor calls and video meetings, customer service teams use recorded speech and some financial institutions incorporate vocal characteristics into identity checks.

As cloning improves, organisations cannot assume that hearing a familiar voice provides meaningful authentication. Payment requests, password resets and confidential instructions need separate verification procedures, particularly where the request involves unusual urgency or attempts to bypass established process.

Synthetic identity is therefore both a legal and operational problem. Rights can support enforcement after misuse, while organisations still need controls that stop an imitation succeeding before lawyers become involved.

Meloni’s EUIPO application may establish another useful layer of protection, or it may reveal the limits of applying trademark doctrine to a new technical capability. Either result would provide a practical test of how far existing intellectual property law can stretch.

Generative AI has made a person’s voice reproducible at negligible marginal cost. The law is now being asked whether that same voice can function as protected intellectual property when the copy becomes almost as easy to create as the original is to record.