In April 2023, the music world was shaken by an unprecedented event.
A song sung with the voices of Drake and The Weeknd garnered over fifteen million views on social media and more than five thousand plays on Spotify, but neither of them had recorded the track.
The song, titled Heart on My Sleeve, had been generated by an artificial intelligence system that replicated their voices with such quality that they were indistinguishable from the originals.
The phenomenon of digital replicas is not new, but it has spread rapidly in recent years, increasing the likelihood of being unknowingly involved in unpleasant situations, not only for artists but widely so.
The harms of digital replicas
According to research conducted mainly in the United States, digital replicas can cause damage classified primarily into three different types.
The first concerns the distribution of sexually explicit content created through deepfakes. A recent study revealed that 98% of deepfake videos online contain explicit images and that 99% of the people represented are women, while the number of cases where students create and distribute fake images of peers and classmates—even in intimate contexts—is growing.
The second danger involves fraud. The ability to replicate voices and images with extreme precision is becoming a powerful weapon in the hands of scammers. Among the most famous cases is the cloning of the voice of a multinational financial CEO to carry out a $25.6 million fraud, and the replication of a son’s voice to extort $9,000 from a father who believed he was helping him pay bail. Promotional use is also frequent, where voices similar to those of famous people are used to advertise products, thus deceiving consumers.
The third, no less significant, danger concerns politics and social participation. Digital replicas are being used to manipulate public opinion and convince people of the truth of events that never happened. Thus, again in the United States, a fake message circulated in which a cloned voice of a Chicago mayoral candidate appeared to justify police brutality, and another message was falsely attributed to the voice of former President Joe Biden, discouraging voter participation in the primaries.
Beyond this general impact, the cloning of voices and physical features has an effect—considered by many to be devastating—in the artistic field, where actors and singers fear that artificial intelligence could completely replace them, as happened with the song Heart on My Sleeve.
Voice or physical appearance, as a natural characteristic of a person, is not protected by copyright, which safeguards original creative works like songs, books, and films, but not personal attributes.
This does not mean it’s impossible to defend oneself, but one must act on a different level—moving from copyright or property law to personality rights, for example by invoking the violation of privacy, image, or misappropriation of identity.
Some U.S. states, like California, explicitly recognize the right to one’s voice as part of the right of publicity, which is a right to the commercial use of one’s image and identity.
This is an entirely different level—not so much economic as involving the inalienable and inviolable rights of the individual, which by their nature are not commercializable.
The NO FAKES Act
It is within this context that a recent legislative proposal should be considered, first introduced in 2023 and now reintroduced in 2025 by a group of bipartisan U.S. senators and representatives. The goal is to establish a sui generis copyright on a person’s voice and physical likeness.
The proposal is called the NO FAKES Act, acronym for “Nurture Originals, Foster Art, and Keep Entertainment Safe Act”, a law aimed at promoting original works, supporting Art, and protecting the entertainment industry.
The idea is to legally introduce an intellectual property right on voice and physical likeness, as a form of property right, transferable to heirs and valid for a person’s entire lifetime and for seventy years after death, and potentially licensable to third parties.
The proposal provides that license agreements must be made in writing, for a maximum duration of ten years, and must specify the specific purposes for which the data will be used. For minors, parents may grant a license for the commercial use of their child’s voice or image, but only for a maximum period of five years and never beyond the child’s eighteenth birthday, subject to court approval.
“Post mortem” transfers or licenses may also be allowed, provided they are made in writing.
Online service providers using unauthorized digital replicas are liable for an offense and subject to heavy penalties unless they can demonstrate that they took all measures to verify the legality of the data, adopted suitable measures to close the accounts of repeat offenders, and, once they received a substantiated and reasoned warning, removed the unauthorized images or digital voices “as soon as technically possible.”
Anyone who uses unauthorized clones in any form, in addition to compensating the affected person for any damages, must pay a penalty of $5,000 per image or voice used if the offender is a natural person, and $25,000 if it is an online service provider. If the provider failed to adopt appropriate measures to reduce the risk of infringement, they may be required to pay a penalty of up to $750,000.
Supporters
The powerful RIAA, Recording Industry Association of America, strongly supports the bill, which also appears to be backed by companies such as OpenAI, Google, Amazon, Adobe, and IBM, while the White House has not yet commented.
Meanwhile, artists seem very satisfied, and one of the first to express appreciation was Randy Travis, who, after suffering a stroke in 2013, made headlines last year for being able to release a new song titled Where That Came From, originally recorded as a demo by another singer, onto which an AI system applied Randy Travis’s voice.
The artist’s wife released a recent statement expressing happiness about being able to record a new song as an extension of his art, appreciating the technical innovation that allows him to express himself, but clarified that his case is very different from what others might do with artificial intelligence by using artists’ voices without their consent, and she expressed support for the bill.
If the NO FAKES Act passes, a person’s voice and image—highly personal data of the individual—will thus become transferable and marketable goods. In Europe, such a possibility might raise eyebrows, but it is not excluded that the same path will be followed.
Those affected by cloning will be able to manage it or be compensated economically, but it will certainly not reduce the risk of fraud or political manipulation, nor will it make the world a more transparent or safer place.