Christodoulides does not rule out AI interference in next Cypriot elections

President Nikos Christodoulides on Friday refused to rule out the prospect of “malevolent actors” using artificial intelligence-based disinformation campaigns to sway voters at the next Cypriot presidential elections, which will take place in February 2028, as he arrived at the day’s informal European Council summit in Dublin. “I will not hide that there is concern about how the technology is being used, especially by malevolent actors, so as to affect electoral processes, the results of elections, and so on in member states of the European Union,” he said. As such, he added, European states must codify “framework which will protect their societies”. Asked specifically about the next elections in Cyprus, he said that “from the moment that it is something which we are seeing present itself in Europe, and we see it in practice, and the Republic of Cyprus is a member state of the European Union, we are of course all vulnerable to this technological threat”. “What is the most important thing, and is something which we discussed yesterday … is not just to analyse these malicious acts, but to have a framework in place so that such malicious acts cannot happen,” he said. He said on this front that there are two “directions” which the EU must take, the first of which is “to inform the public”, while the other is that “as governments, we have a responsibility to have the framework in force so that such things cannot happen”. Commission plans to address AI electoral interference These comments come with the European Commission having on Tuesday launched a consultation on planned revisions to the EU’s digital services act, which focus on what it describes as “very large online platforms” and “very large online search engines”, and planned measures to address “information manipulation risks to electoral processes”. The commission’s planned revisions to the law include stipulations that online platforms “detect coordinated inauthentic behaviour earlier” and “dismantle the networks behind it”, while also being expected to “spot and assess suspicious trends before they go viral”. Most importantly, it says that online platforms available in Europe would, were the revisions to pass into law, “test their algorithms for manipulation before every national election” and “keep a record of the changes made to their recommender systems in the six months before a vote”. On the matter of artificial intelligence, it says that “deepfakes should be clearly labelled and keep their label when shared”, and that “illegal deepfakes of [electoral] candidates should come down fast”. It also says that following any election, those platforms should then report to the European Commission within 30 days “on how they mitigated risks”, while also publishing a “public version” of those reports. The consultation process will end on November 23, after which it is expected that the commission will submit the proposed amendments to the European Parliament and the Council of the EU.