|
Getting your Trinity Audio player ready...
|
From a fake “AI band” in Moldova’s 2025 selection to San Marino’s experiment with AI-assisted songwriting, artificial intelligence is already challenging how music is made and judged. Eurovision’s organisers have discussed the problem, but as of May 2025 they have not clearly explained how AI-assisted entries should be declared, assessed or credited. Will the contest’s human spirit survive?
The Moldova hoax
In January 2025, Moldova’s national selection erupted into scandal when Eblansh Band withdrew, officially citing a member’s illness. The act had been listed among the participants in Etapa Națională with the song “O, Moldova mea dragă!” (“Oh, my dear Moldova!”).
Moldovan and regional reports subsequently described Eblansh Band as fictitious and the song as an AI-generated prank attributed to Ukrainian pranksters. That attribution should be treated cautiously: the available reporting largely repeated online claims rather than documenting a formal investigation or naming the people behind it.
What is clear is that the entry passed the administrative screening stage. Daniela Crudu, Moldova’s Eurovision head of delegation, said the submissions had not been checked for AI at that point, with the live jury auditions expected to assess the entries instead. Eblansh Band withdrew before it reached that stage.
That exposed a genuine vulnerability in the selection process. A fictional act and an AI-associated song could enter the official lineup without anyone apparently being required to declare how the music had been made. It did not, however, prove that an AI song could have reached the Eurovision stage.
Moldova withdrew from Eurovision 2025 on 22 January, four days after the live audition round. Moldova 1 director Corneliu Durnescu and Crudu cited economic, administrative and artistic difficulties, including concerns about public interest and the quality of the national selection. The withdrawal was not officially attributed solely, or specifically, to the Eblansh incident.
San Marino goes AI-assisted
San Marino took a more deliberate approach. Its San Marino Sessions was an experimental collaboration between SMRTV and London-based music-technology start-up Casperaki, linked to the 2024 edition of Una voce per San Marino.
But this was not four entirely machine-composed songs simply dropped into a national final. Applicants submitted material through Casperaki’s AI-assisted platform. Ten proposals were then selected for a songwriting camp in London, involving human songwriters, producers and artists. Four completed songs were shortlisted, and one became a finalist in Una voce per San Marino.
It did not win. Spanish band Megara won the 24 February 2024 final with “11:11”. So the safer description is “AI-assisted” or “AI-developed with human collaborators”, rather than “four songs composed by an algorithm”. Even so, if one microstate can publicly experiment with human–AI songwriting, what about the rest of Europe?
EBU’s AI dilemma
Behind both stories lies a glaring gap: Eurovision’s public-facing rules and procedures had not clearly explained how to handle AI-written or AI-assisted entries.
In 2023, EBU deputy director general and media director Jean-Philip De Tender discussed the possibility of AI-generated songs entering Eurovision and said the organisation was reflecting on how the issue might be handled. Contemporary reports presented his concern that Eurovision creativity should come from humans rather than machines.
That was a public discussion, not evidence of a comprehensive AI ban. Research available by May 2025 does not establish that the EBU had adopted a dedicated AI policy for the 2024 or 2025 contests. De Tender’s later remarks also offered a more positive view of AI in public-service media, stressing transparency about AI use and the importance of human intervention.
That distinction matters. “AI is involved” is not the same thing as “a machine wrote the entire song”. Copyright, authorship, performer eligibility and disclosure are separate questions. An AI-assisted song is not automatically ownerless or ineligible; the answer depends on the human contribution and the rules that apply.
When an AI song tops the charts
Beyond Eurovision, AI-generated music is already being discussed as part of digital culture. A song titled “YAJU&U” has been cited online as an AI-generated Japanese hit, including claims that it reached number one on Spotify Japan’s Viral 50 playlist.
But those claims are not adequately supported by reliable contemporary music reporting or historical chart data. So it would be premature to call the track one of 2025’s biggest domestic hits. The broader question remains valid, though: might listeners accept AI involvement when it is disclosed, rather than when it is quietly passed off as conventional human creation?
Authenticity versus innovation
The rise of AI has split the music community. In April 2024, more than 200 artists signed an Artist Rights Alliance open letter urging AI developers, technology companies, platforms and digital music services not to use AI to undermine or replace human artistry. Signatories included Billie Eilish, Katy Perry, Nicki Minaj, Jon Bon Jovi and Stevie Wonder.
The letter was not a blanket rejection of every use of AI. Its concerns included unauthorised training, imitation of voices and likenesses, the replacement of human artists and the question of compensation. In other words, the argument is not simply “all technology is evil”. It is about who controls it, who profits and whether human creators are being pushed out.
That clash highlights a core question: is it better to create songs by committee, with partial creative credit given to a machine, or to keep the emotional story firmly attached to human performers and writers?
Boot camp generation
Eurovision already relies on professional songwriting camps where many writers collaborate on a single entry. If a pop song with up to ten human co-writers can still feel genuine to fans, is an AI-assisted tune truly less “authentic” than a track from a cookie-cutter pop factory?
If Eurovision fans have accepted radio-friendly formula pop from committees, why should AI be any more suspect?
Of course, the difference is that an AI isn’t human – it has no story or soul.
This is where the ethical concerns arise. Who “owns” an AI-written song? Does the composer of the algorithm get royalties, or the user who prompted it? And more philosophically: what happens to the emotional core of a ballad if no human heart beat behind it?
Eurovision thrives on moments of genuine feeling. If those messages are spat out by a neural network, do they still resonate?
Music’s soul at risk
The fear is that overuse of AI could commodify songwriting. Instead of artists mining personal experiences, machine-generated formulas might push homogenised pop. Lyrics become patterns generated from data rather than sentiments lived by a person.
This risks stripping away the spontaneity and quirkiness that gave Eurovision gems like “Fairytale” or “Madness of Love” their magic.
The EBU and national broadcasters need to consider clarifying the rules: perhaps requiring declarations of AI use in submitted songs, or setting boundaries around machine-generated content. The Moldova episode showed that preliminary screening can miss the issue; San Marino showed that AI can also enter through an openly collaborative process.
Simply sweeping AI under the carpet may no longer be an option.
Above all, this needs a public conversation. Broadcasters, songwriters and fans should debate what Eurovision represents. Do we want a contest where machines can buy the spotlight, or one where human stories and imperfections still shine?
After all, Eurovision’s motto is “Unite by Music,” not “Unite by Machine.”
