An AIgenerated video of London in decline, shared on the social media site X account of British farright activist Tommy Robinson, is viewed on a smart phone in London. — © AFP HENRY NICHOLLS
Anna MALPAS
Daubed in Arabiclooking graffiti, London’s Big Ben is shown smouldering above piles of rubbish and crowds dressed in traditional Islamic garb in an AIgenerated, dystopian vision of the British capital.
Farright leaders and politicians are seizing on such clips of reimagined European cities changed by migration to promote racist views, falsely suggesting AI is objectively predicting the future.
The videos — which show immigrants “replacing” white people — can be made quickly using popular chatbots, despite guardrails intended to block harmful content, experts told AFP.
“AI tools are being exploited to visualise and spread extremist narratives,” the CEO of the Center for Countering Digital Hate watchdog, Imran Ahmed, told AFP.
British farright leader Tommy Robinson in June reposted the video of “London in 2050” on X, gaining over half a million views.
“Europe in general is doomed,” one viewer responded.
Robinson — who has posted similar AI videos of New York, Milan and Brussels — led the largest farright march in central London for many years in September, when up to 150,000 people demonstrated against the influx of migrants.
“Moderation systems are consistently failing across all platforms to prevent this content from being created and shared,” said Ahmed of the Center for Countering Digital Hate.
He singled out X, owned by tech billionaire Elon Musk, as “very powerful for amplifying hate and disinformation”.
TikTok has banned the creator account behind the videos posted by Robinson. According to the platform, it bans accounts that repeatedly promote hateful ideology, including conspiracy theories.
But such videos have gained millions of views across social media and have been reposted by Austrian radical nationalist Martin Sellner and Belgian rightwing parliamentarian Sam van Rooy.
Italian MEP Silvia Sardone from rightwing populist party Lega in April posted a dystopian video of Milan on Facebook, asking whether “we really want this future.”
Dutch farright leader Geert Wilders’ Party for Freedom released an AI video of women in Muslim headscarves for the October elections titled “Netherlands in 2050.”
He has predicted that Islam will be the Netherlands’ largest religion by that time, despite just six percent of the population identifying as Muslim.
Such videos amplify “harmful stereotypes… that can fuel violence”, said Beatriz Lopes Buarque, an academic at the London School of Economics researching digital politics and conspiracy theories.
“Mass radicalisation facilitated by AI is getting worse,” she told AFP.
– ‘Hate is profitable’ –
Using a pseudonym, the creator of the videos reposted by Robinson offers paid courses to teach people how to make their own AI clips, suggesting “conspiracy theories” make a “great” topic to attract clicks.
“The problem is that now we live in a society in which hate is very profitable,” Buarque said.
Racist video creators appear to be based in various countries including Greece and Britain, although they hide their locations.
Their videos are a “visual representation of the great replacement conspiracy theory,” Buarque said.
Popularised by a French writer, this claims Western elites are complicit in eradicating the local population and “replacing” them with immigrants.
“This particular conspiracy theory has often been mentioned as a justification for terrorist attacks,” said Buarque.
Round dates such as 2050 also crop up in a similar “white genocide” conspiracy theory, which has antiSemitic elements, she added.
AFP digital reporters in Europe asked ChatGPT, GROK, Gemini and VEO 3 to show London and other cities in 2050, but found this generally generated positive images.
Experts, however, said chatbots could be easily guided to create racist images.
None has moderation that “is 100 percent accurate”, said Salvatore Romano, head of research at AI Forensics.
“This… leaves the space for malicious actors to exploit chatbots to produce images like the ones on migrants.”
Marc Owen Jones, an academic specialising in disinformation at Northwestern University’s Qatar campus, found ChatGPT refused to show ethnic groups “in degrading, stereotypical, or dehumanising ways”.
But it agreed to visualise “a bleak, diverse, survivalist London” and then make it “more inclusive, with mosques too”.
The final image shows bearded, ragged men rowing on a rubbishstrewn River Thames, with mosques dominating the skyline.
AFP, along with more than 100 other factchecking organisations, is paid by TikTok and Facebook parent Meta to verify videos that potentially contain false information.
