No Author, Many Violations: AI-Generated Music and the Limits of Copyright and Personality Rights

Author: Luiza Rey

Over the past weeks, many people have come across short videos and posts repeatedly using the song “A Sina de Ofélia”. 

It sounds familiar, emotional, and unusually convincing. 

What is not immediately obvious is that this song was entirely generated by artificial intelligence. The Portuguese lyrics, the musical arrangement, and most importantly, the voice, were all created synthetically.

That voice closely imitates the Brazilian singer Luísa Sonza, without her authorization, to perform a translated version of a song originally released by Taylor Swift. 

The viral success of the track exposes a structural problem that goes well beyond music fandom or internet novelty. 

In 2025, Taylor Swift released “The Fate of Ophelia”, a song inspired by Ophelia, the tragic character from Hamlet, by William Shakespeare. 

In Shakespeare’s play, Ophelia is driven into madness by loss and emotional betrayal and ultimately drowns. Swift reimagines this figure as a metaphor for isolation, heartbreak, and the conscious decision to step away from a self-destructive spiral after being treated with care and respect. 

Shortly after the release of the original song, a Portuguese version titled “A Sina de Ofélia” began circulating online. 

It was not recorded by any human performer. It was generated. 

Behind this type of production is a class of systems known as Generative Adversarial Networks (GANs). 

These systems rely on two AI models trained in opposition: a generator that produces content and a discriminator that evaluates how closely that content resembles real-world data. The generator improves by repeatedly failing the discriminator’s tests. 

In this case, the system was trained on extensive recordings of the Brazilian artist’s voice, learning tone, cadence, phrasing, and stylistic patterns. Over time, the synthetic output became convincing enough not only to satisfy the algorithm, but also to mislead human listeners. 

This single example already reveals a dense cluster of legal problems: 

• unauthorized creation of a derivative work 

• infringement of copyright and related rights 

• violation of personality and voice rights 

• reputational and economic harm 

• confusion as to authorship and authenticity 

• deception of the public 

The original song was translated and adapted, which legally qualifies as a derivative work. Under copyright law, derivative works require prior authorization from the original rights holders, regardless of whether the use is commercial. 

At the same time, the use of an identifiable voice, even through AI-generated imitation, implicates personality rights. Voice, like name and image, is a legally protected marker of identity in many jurisdictions. 

There is a deeper paradox at play. 

If a work is produced entirely by AI, without meaningful human creative input, it generally does not qualify for copyright protection. Copyright law remains anchored in human authorship. 

As a result, works like “A Sina de Ofélia” exist in a legal grey zone. They may attract massive public attention, yet lack a legitimate author or rights holder. They cannot be licensed, properly monetized, or defended as original works, while still infringing the rights of others. 

This explains why platforms removed the track quickly. There was no valid license, no identifiable rights holder, and clear exposure to third-party claims. 

From a U.S. legal perspective, this case aligns closely with ongoing disputes around AI, copyright, and identity. 

The U.S. Copyright Office has stated that works generated exclusively by artificial intelligence, without human creative contribution, are not eligible for copyright protection. At the same time, such works may still infringe existing copyrights if they reproduce recognizable elements of protected material. 

This creates a practical enforcement asymmetry. Rights holders can rely on DMCA takedown mechanisms to remove infringing AI-generated content from platforms like YouTube, Spotify, and TikTok, often without judicial involvement. This approach has already been used extensively in cases involving AI-generated songs imitating well-known artists. 

Fair use arguments are unlikely to succeed when the output is highly similar, commercially substitutable, or deliberately designed to sound like a specific artist. The closer the imitation, the weaker the defense. 

From a European Union perspective, the legal analysis is broader and, in some respects, stricter. 

EU copyright law gives authors exclusive control over adaptations and translations, meaning that unauthorized AI-generated versions qualify as infringing derivative works. In parallel, European legal systems place strong emphasis on personality rights, including protection against the unauthorized exploitation of a person’s identity or voice, particularly where there is a risk of public confusion or parasitic advantage. 

Importantly, the EU framework also places increasing responsibility on platforms. Under the Digital Services Act and related legislation, large online platforms are subject to enhanced duties of diligence, transparency in content moderation, and mitigation of systemic risks, including misleading or deceptive synthetic media. Once notified, platforms are expected not only to remove infringing content, but also to take reasonable steps to prevent its reappearance. 

In practice, this means that AI-generated music using identifiable voices without authorization is not only a private dispute between artists and creators. It becomes a compliance issue for platforms themselves, with potential regulatory consequences if patterns of abuse are not addressed. 

The absence of comprehensive AI regulation does not imply legal tolerance. 

Several pressure points are becoming increasingly visible: 

• explicit consent requirements at the data collection and training stage 

• traceability and auditability of training datasets 

• mandatory labeling of synthetic content 

• licensing requirements for the use of identifiable voices and likenesses 

• shared liability between content creators and distribution platforms 

These discussions are already shaping regulatory agendas in both the EU and the United States. 

“A Sina de Ofélia” is not an isolated anomaly. 

It is a preview. 

Generative AI can now produce content that feels authentic, emotionally resonant, and culturally embedded. But legal systems still depend on concepts of authorship, consent, and accountability that machines cannot satisfy on their own. 

AI does not eliminate legal boundaries. It tests them. 

The real question is not whether synthetic creativity will continue to expand, but whether law, platforms, and creators will adapt fast enough to preserve authorship, identity, and public trust.

By: Luiza Rey