Political campaigning in times of AI

The 2016 election campaign in the United States caught us off guard. Who could have anticipated that the candidacy of a ridiculous media millionaire who silenced vedetes and asked for birth certificates would converge with a savage operation of Russian influence and a handful of Macedonian mushroom peddling villains in an underground program capable of distorting half of Americans without the other half knowing. We didn't know about Facebook's new tool to find susceptible voters and make them available to their customers. We didn't understand the radicalizing potential of YouTube. We didn't know what Cambridge Analytica had done for the Brexit campaign that summer. Little did we know that someone could create tens of thousands of fake accounts to poison tens of thousands of people with false narratives about their own communities. The anti-vaxxers were poor ignorant peasants. QAnon did not exist yet.

There are now university departments around the world dedicated to researching computational propaganda. There are books analyzing the disinformation machine in Russia, China, Iran, and Israel. There are workshops to detect "coordinated inauthentic behaviors" and a planetary network of fact-checkers with enough experience to know where fake news is created and how it is distributed, translated and dispersed throughout the local chapter. And we know that all the operations we know will be assisted by artificial intelligence models capable of creating people, documents and events that do not exist or have not taken place. But we have done nothing to prepare. It's still 2016.

We will again cover only the visible part of the political campaign. The one that appears on the marquees, the one that is shared on Twitter and seen on television. Even knowing that the real campaign is many different campaigns that choose their recipients very well. We will continue to deny fake news that has already reached the pockets of people who are not subscribed to any newspaper through encrypted messaging platforms such as Telegram or WhatsApp groups. There is no effective campaign to help citizens identify dark campaigns, regardless of origin. We have not constituted an electoral security team that coordinates the efforts of universities, research centers, platforms, operators, state security forces and the media to guarantee the honorability of the democratic process.

We will protest on Twitter when the platforms fail to comply with the day of reflection contained in the Organic Law of the General Electoral Regime. No one expects debates to be conducted with honour. There will be no bipartisan alliance to jointly monitor influence campaigns. But we know there will be messages designed to discredit the recount and delegitimize the electoral process and its results. We know everything and we have done nothing. It's still 2016.

This is the year of generative AI models. The campaign is more agile, cheaper, more prolific than ever. Distribution networks have been consolidated. You no longer need audiovisual production to make deepfakes. We are not prepared for the upcoming campaign, but this time not for lack of information.

Subscribe to continue reading

Read without limits

Read more

I'm already a subscriber