Last April, as the campaign for the European elections was in full swing, the French far right enjoyed massive online support from two very popular influencers on TikTok: Amandine Le Pen, niece of Marine Le Pen, of the National Rally, and Léna Maréchal, niece of Marion Maréchal, candidate of Eric Zemmour’s Reconquête party. With tens of thousands of subscribers and millions of views, these influencers, sometimes posing in a lascivious and suggestive way, shared content encouraging people to vote for the far right in the elections.
The far right has long demonstrated, much more than the left, its mastery of the codes of social networks and the aesthetics of online pop culture. Problem: these two young women do not exist. They are products of artificial intelligence (AI), deepfakes imitating real people almost perfectly. This use of artificial intelligence to support political groups, whether radical or more traditional (the authors of the aforementioned creations are, moreover, difficult to identify), is however only the tip of the iceberg. Below, under the surface, violent extremists are already exploiting the unprecedented possibilities offered by generative AI to serve their cause.
Far-right groups are using AI to create large-scale disinformation campaigns by producing visual content inspired by the aesthetics of Nazism. Thanks to its generative capacity, they gather information on the use and construction of explosive devices, or even replicate the work of graphic designers of well-known propagandists in the most radical circles, such as the Canadian Dark Foreigner (who notably popularized the bloody aesthetics of far-right accelerationist terrorist groups), and do so with much greater ease and speed than before.
In France, during the last legislative elections, the song I won’t leavecreated by an AI, spread xenophobic and racist messages on a large scale thanks to its distribution by radical right accounts, garnering millions of views.
Several AI music production sites have recently been used to produce songs glorifying ISIS or to promote anti-Semitic conspiracy theories. On Telegram, radical Islamist groups are using generative AI to mass-produce propaganda and translate it for non-Arabic speakers in record time.
Generative AI is still in its infancy. Some researchers believe that we are experiencing a shock similar to the arrival of the Internet, and the impact will be absolutely spectacular in our daily lives over the coming years and decades. AI does not just produce: it short-circuits the entire chain of intermediaries formerly placed between individual ideation and final content. In this way, generative AI pushes to the paroxysm the transformation of the individual into prosumer (contraction of the words ” producer ” And ” consumer “). Social networks, which are known to promote the virality of content that provokes strong emotions, such as laughter, fear or terror, then allow this content to be disseminated on a large scale.
Extremist groups have already seized this opportunity to pursue their political and ideological goals, despite measures put in place by digital platforms to limit violent or hateful content, or prevent generative AI from being used for these purposes.
The real danger is there: the almost unstoppable spread, on a scale never seen before, of extremist content designed to evade any legislation or moderation. And the risk associated with the proliferation of this content is great: massive and more real-than-life disinformation campaigns; amplification of hate speech online (which is known to de facto reinforce discrimination and negative attitudes towards the targeted groups); propaganda that is aestheticized and crafted in such a way as to promote the penetration of the discourse on social networks; and increased capacity for radicalization, interference or destabilization.
From visual propaganda to deepfakesgenerative AI will make these contents ever more realistic, ever more difficult to detect, ever more difficult to stop, and therefore ever more difficult to combat. Their destructive potential, online and offline, will be increased, imposing the need for real reflection on the supervision, ethics and regulation of AI. Because it is illusory to think that the physical universe is detached from cyberspace: it constitutes a layer in its own right. And generative AI has just completed this fusion.