The answer is already there: as soon as President Biden announced his candidacy for re-election, the Republican camp flooded the networks with true-to-life apocalyptic videos of migrants pouring in, soldiers occupying the streets and Taiwan being bombed. This clip, generated by artificial intelligence, provoked a great deal of emotion and comment - and rightly so - on the confusion between fiction and reality in an election campaign. In truth, this debate is one of the oldest in the world, since it concerns honesty or lies in politics. Voters didn't wait for this type of content to cast doubt on everything that came from the other side, and give their uncritical support to everything that came from their own side.
No, the real game-changer with the advent of AI is the industrialisation of ultra-targeting of voters. As the Cambdrige Analytica scandal revealed, social networks had already made it possible to push messages tailored to each audience, via the aspiration of personal data. In a world where it is possible to generate virtually bespoke content for each user, we are in for a Cambridge Analytica on steroids. And in this new game, we all know that the extremes have an advantage because of the very nature of the human feelings they rely on to thrive. So how can we prepare? How do we deal with it?
There are three possible scenarios. The first is that of unbridled AI, without any form of regulation or brake. The result is a total cacophony, with voters harassed by an unprecedented amount of false information thrown around in far-fetched and crude scenarios. It's a zero-sum game that adds to the weariness and disinterest in the democratic game, freezing it in place with existing trends and making it impossible for a new political offering to emerge.
The second is, of course, the need for strict regulations governing the use of AI in election campaigns, similar to those we have in France for electoral propaganda. We could define a pre-election period during which the use of AI is blocked, particularly by publishers, for all political content. A vast administrative and technical machine would be needed, for an uncertain result. Who would be legitimate in France to lead such a battle? The Commission nationale des comptes de campagne et des financements politiques (CNCCFP)? The Commission nationale de l'informatique et des libertés (CNIL)? The Council of State? And what about the European level, in the run-up to the 2024 elections?
The last option is that of self-limitation, by obtaining a commitment from all parties that they will not use AI in their electoral practices. It is not a question of relying on the goodwill and good faith of all the parties, but rather of making opacity on the subject politically costly. Why not imagine a charter, a manifesto signed by the entire political spectrum? Faced with the fear of seeing public debate poisoned by suspicion of AI, those who are absent would quickly lose out.
Ultimately, to preserve democracy, we need to invent a genuine safe place. A new digital agora that gives us oxygen, in the low-tech spirit that made the early Internet a space of infinite possibilities and riches. A kind of cyber sanctuary protected from the excesses of digital cacophony... And one that could find its counterpart in the physical world, since the theatre of democracy must remain the public space.
Ultimately, if we are to preserve democracy, we will have to invent a genuine safe place. A new digital agora that gives us oxygen, in the low-tech spirit that made the early Internet a space of infinite possibilities and riches. A kind of cyber sanctuary protected from the excesses of digital cacophony... And one that could find its counterpart in the physical world, since the theatre of democracy must remain the public space. Right now, we need to rethink the use of public buildings in this new electoral context, to ensure that all areas of the country are able to take part in a complex, adversarial debate based on everyone's arguments. The equivalent of the War Room concept, where information is used to understand the unfolding crisis, applied this time to public debate, in the service of informing citizens. It's a conception of politics that goes against the grain of automatically generated content, automatically repeated by an automatically targeted profile.