With alarmingly racist lyrics and an undeniably catchy chorus, the song ‘Je Partira Pas’ spread like wildfire on French social media amid the 2024 French elections – elections that saw the popularity of far-right parties surge. The viral song was out of the ordinary in that it was generated by AI, part of a growing trend of the far-right adopting new technologies to spread their views.
While AI-generated deepfake images and audio circulating disinformation have made headlines related to elections across Europe and the US, racist AI songs have also been generated at a significant scale. Far-right and white supremacist AI-generated songs have been created in response to recent events, from the UK’s anti-migrant riots in Southport, to the rise of US Democratic candidate Kamala Harris to the victory of Donald Trump.
In this article, ArtsHub speaks with far-right experts, while presenting its own investigation into the prominent generative AI music platforms of Udio and Suno AI.
In this article:
History of far-right music
Those on the far-right have not only long been avid users of various art forms, including music, but also historically have been quick adopters of the newest trends and technologies.
Helen Young, a researcher at AVERT, specialising in far-right popular culture and white supremacy, explains that the far-right is generally good at utilising popular culture to make “a connection with people who share an interest – whether it’s punk music or metal music. Certainly, the far-right are at large with all different kinds of culture, films, books, TV, all kinds of stuff”.
Young says that music aids the far-right in building a community of people that share extremist views or even drawing in those who “might be a little bit sympathetic even if they don’t start with the extreme beliefs”.
It then becomes not just “something that a person may think or feel something about but … part of all the different parts of their life … [and] recreation, like listening to music, a thing that they go and celebrate at a white power festival or a white power gig. So, it’s partly about recruiting and partly about community building, once people start to come together with those shared hateful beliefs,” she says.
Liam Gillespie, an expert in criminology who has conducted research into the far-right’s use of sound, agrees. “Music has been implemented as a narrative and framing tool, and for recruitment and political messaging,” he says.
“Listening to and enjoying music is a way ‘in’ to a group’s ideology. Actually, attending concerts/venues where those bands are playing is also a social activity, a ritual that members can undertake to create and/or strengthen the bond between them,” he continues.
Imogen Richards, another expert in criminology and researcher for AVERT, specifically delves into the far-right movement’s syncretic nature. This means that they often superficially adopt and utilise elements from different movements and aesthetics like culture “in order to appeal to a diverse demographic and to basically portray themselves as something new,” she explains.
For example, “KKK music of the 1950/60s incorporated hatenanny records,” she says. This played upon the progressive folk music movement of Hootenanny. Punk music and Oi! music of the 70s and 80s, while popularly associated with the far-right, actually predated Neo-Nazis and white supremacists, who appropriated the music genre “when they saw the popularity of the rise of punk rock,” says Richards.
“Some former Neo-Nazis … have highlighted that the punk rock music scene was partly entered into by white supremacists and national socialist activists for the purpose of seeking out fights with progressive anti-racist punks … but also in appealing to younger generations.”
In fact, the Rock Against Communism white power concerts in the 1970s were a direct reaction against progressive, anti-fascist movements such as Rock Against Racism. The same syncretism can also be seen in the far-right’s adoption of neo-folk, as well as national socialist black metal from the 1980s and 90s, Richards notes.
This has continued in the modern day. “This repurposing has carried over into the latest kind of iterations, which include things like fashwave, schizo wave and Nazi hip hop,” says Richards, which are also “trying to leverage the countercultural appeal of these pre-existing trends and styles”.
Those particular music styles are seeing increasing AI generation in general, so it “makes sense that would carry over into the … white supremacist music scene as well,” adds Richards. Further, the ease of creating AI music means that “people who don’t already have the technical skills, abilities or knowledge to be able to produce the music” can do so through AI technology.
“Like the further right elements of the Trump campaign tapping into a culturally popular aesthetic and sentiment … now we’re just seeing that happen in more technologically sophisticated ways and in ways that are transnational, so they can also draw on support from groups internationally.
“We see the AI-generated images, music and aesthetics more broadly being used to scoop up and attract those audiences en masse,” Richards continues.
One example is a parody of Eurovision that presents Qatar as a participant, with an AI-generated xenophobic song that blames the country for the Islamisation of Europe, “causing a mess” and linking it to terrorist groups.
Young summarises that the far-right is “manifesting in new ways, but not very much is really new… The music things have been around for a long time. Now they’re using AI, but they’ve been using music for 40 to 50 years. They’ve been using books for that long too. They’ve been exploiting mainstream media for a long time,” she says.
Prominent AI music examples
AI far-right music has been especially prominent around elections, going viral as well as drawing support towards far-right political candidates or echoing anti-migrant rhetoric.
In France, the song ‘Je Partira Pas (You Will Leave)’ espousing anti-migrant racist messaging went viral across social media in the midst of the country’s elections. Calling for deportations, the song echoed the views of the French far-right National Rally party. Led by Jordan Bardella, the party almost came to power with its anti-migrant stance. One lyric reads: “When Bardella comes, you’ll go home. You’ll wear your djellaba. You can pray all day. Now you’re starting to annoy us.”
Shared across YouTube, Twitter, TikTok and several other mainstream platforms, the AI-generated song has been used to accompany everything from AI-generated music videos to dances, far-right rally footage and multilingual translations.
Likewise, in Germany, AI songs calling for the deportation of immigrants have been generated by extreme far-right political groups such as the AfD (Alternative for Germany), which saw success in several eastern German states amidst the country’s elections. The ‘Remigration Hit’ is a dance track made by AfD’s youth wing, featuring an AI-generated video of migrants being deported while blond-haired and blue-eyed people celebrate. AI video generators such as Luma are becoming more prominent for creating music videos, with Google recently releasing a new video generator.
Another AI-generated song reached the top 50 on the German music charts in August 2024. This one featured the derogatory term ‘talahon’ – frequently used by far-right groups to describe young men with migrant backgrounds. The Udio generated song went viral across TikTok and Spotify.
Similar patterns have been seen across Europe where far-right parties have utilised innovative new tactics while also seeing waves of support, including most recently in Romania where a far-right party has been gaining prominence.
Other far-right songs utilise AI to mimic the voices of certain far-right figures in politics, often using software such as Covers.ai or Voicify AI, which allow users to make songs with certain voices. For example, the AI-generated rap song ‘First Day Out’, using Trump’s voice, was so popular that it hit number two on iTunes’ rap chart in mid-2023. The song was also promoted on YouTube by several right-wing personalities and politicians such as Benny Johnson and Candace Owens.
Over in Asia, the Indian elections also saw the rise of far-right Hindutva music spreading hatred towards Muslims. A plethora of these songs have been generated by AI, with ones mimicking Prime Minister Narendra Modi’s voice becoming particularly popular. These include examples such as ‘Babam Bam’ sung by an AI-generated Modi, which gained millions of views in the form of reels or videos. AI is also being used to create songs out of speeches made by Modi’s Hindu nationalist party.
Similar AI songs have arisen in France, including an AI-generated clip that overlaid a video of Celine Dion with an AI generated song supporting Bardella, which went viral on TikTok, as revealed by an investigation by the Digital Forensic Research Lab. Another uses Taylor Swift’s voice to encourage violence against Black people, which was shared on Telegram according to DTTM. In another video, AI is used to generate Martin Luther King’s voice in a speech supporting Trump and rhythmically synchronised to music. Similar Hitler speeches synchronised to music are also prevalent.
Investigation into AI music generators and the far-right
In an innovative tactic to more quickly spread their extremist messages, those on the far-right have utilised AI music generators, particularly Udio and Suno AI, to create songs. This is due to the accessibility of such platforms, making it very easy to create an account.
Earlier this year, the Anti-Defamation League also documented various other racist and white-power AI-generated songs, providing various policy recommendations. Since then, this content has only proliferated, as they are very publicly available and easy to find.
In fact, in an ArtsHub investigation conducted since August, typing in certain keywords on Udio and Suno AI related to anti-migrant rhetoric, or recent far-right related events, brings up AI-generated hate songs. For example, searching the words ‘deport’ into both Udio and Suno immediately leads to a wide range of songs that target minority groups, from Muslims to Asians to Africans. One notable result is ‘Racist Anthem’, which repeatedly demonises Asians.
Moreover, typing keywords such as ‘Southport’ in both platforms brings up a large number of AI songs spreading disinformation and falsely accusing Muslims of causing violence in the UK. A pattern of AI-generated songs proliferating occurs with other far-right related events such as Donald Trump accusing Haitian migrants of eating pets earlier this year. A song called ‘Hero of Springfield’, about Donald Trump saving cats and dogs from Haitian migrants in Ohio, is present on Udio.
AI-generated songs created during Kamala Harris’ rise as the Democrat presidential candidate often criticised her, based on her Indian and Jamaican heritage. One such song, since removed, called ‘Kamala Harris Career (it makes fun of her)’ criticised her multiethnic identity. Another song, ‘Kamala Coconut’ on Suno makes a comparison between a coconut and her face.
Tracing these songs on social media platforms such as Gab AI, YouTube and TikTok, ArtsHub found various users proudly sharing white power music. For example, there is a channel called Wanderer’s Choice on Gab AI and New Rebel Records on X as well as on YouTube, which has been sharing white power songs.
Often these users will also provide tips and techniques on how to create these AI songs, especially on far-right channels like Gab AI, but also on mainstream social media platforms like X, such as in the following posts by xGame and other profiles.
According to another investigation by Active Fence, far-right users have been able to create these songs through custom lyrics, despite Udio and Suno AI having filters. They get around the filters by using deliberate phonetic misspellings of offensive terms, such as using ‘jooz’ instead of Jews. Another ADL investigation discovered a user who bragged about bypassing restrictions through tricking the software to create a song about ‘white power’ by making a prompt for another unrelated topic, but with the title ‘white power’.
A similar tactic used by the far-right is the ‘grandma loophole’ wherein users bypass content filters by asking AI generators to pretend to be their dead grandma.
With these tactics still viable and the number of AI-generated hate songs only increasing, ArtsHub contacted both Suno AI and Udio for a response, but had received no comment by the time of publication.
In the second part of this two-part article, which will be published on Friday 20 December, ArtsHub examines how far-right AI-generated music has spread across Australia and beyond.
