Generation Z's young voters rely heavily on the internet to gain knowledge about local and international politics. However, with the rapid advancement of generative artificial intelligence (GenAI), it has become much easier to create content that appears to be legitimate but is not.

The term "Deepfake" refers to a process that utilises GenAI to generate or modify video and audio content, making it seem like a person is saying or doing something they have not done or said. The phrase is derived from the combination of 'deep learning', which is the model for how GenAI functions, and 'fake', which means 'false'.

Deepfake videos serve various sinister purposes. They can be employed to create fabricated "nude" images of public figures or private citizens. Additionally, they may be used in extortion attempts, where perpetrators falsely claim to have kidnapped family members. Furthermore, deepfake technology can manipulate speeches, interviews, and other media to deceive the public. A notable recent instance of this occurred during the Slovakian parliamentary elections.

Two days before the Slovak elections, a fake audio recording appeared on Facebook. The recording featured Michal Šimečka, the leader of the liberal Progressive Slovakia party, and Monika Tódová, a journalist at the newspaper Denník N. In the recording, they were discussing ways to manipulate the outcome of parliamentary elections, including by buying votes from the marginalised Roma minority. Despite the fact the recording was proven to be a deepfake, the election was won by Robert Fico, a pro-Russian politician who is seen as an ally of Vladimir Putin.

Marian Andrei, host and producer of the TV show 'I Like IT', gave an interview to stirileprotv.ro, a partner of Vibez.pl, about the dangers of deepfake technology in the context of multiple European elections.

"The clip was an audio recording, which made it easier for the perpetrators to create a deepfake and much more difficult to detect. The audio was released during the 48 hours when politicians are prohibited from communicating before an election. The candidate was unable to refute the fake clip without violating election laws. Consequently, he lost the election despite being the frontrunner in the polls by a margin of approximately six to seven per cent," Andrei explained.

Andrei warned that thanks to GenAI, creating this kind of content has become incredibly easy, even for those without experience with the technology. "I am aware that deepfakes are being sold on the darknet for a relatively low price. It's concerning to know that Telegram channels that offer deepfake services can be easily found on Google. Unfortunately, it's that accessible and uncomplicated," he stated.

The expert advised being extremely cautious, particularly with videos that have a low resolution and unsynchronised audio with the movement of the lips. As he previously mentioned, voice deepfakes are challenging to detect, particularly for people who are not experts in the field. Moreover, GenAI is becoming increasingly skilled, and soon, distinguishing between generated and genuine recordings (either audio or video) will be almost impossible.

Andrei stressed that he believes the platforms where deepfakes are distributed are not interested in verifying and controlling them. "I haven't seen any deepfake clips appear organically on my channels. They were all sponsored videos," he admitted.

The European Union plans to implement the AI Act in April to regulate all aspects related to artificial intelligence. The new laws will be fully enforceable 24 months after the directive comes into effect, while certain parts, such as the transparency obligation, will take effect after one year of passing the AI Act.

As we approach this year's elections, it is crucial to be cautious about sensationalist videos and audio recordings. Verifying all information and materials found on social media with reliable news sources is advisable.

Source: stirileprotv.ro, europarl.europa.eu