Seoul, 7 May. (askanews) – In a classic office south of Seoul, a team of around 20 employees is fighting an increasingly difficult battle: monitoring content generated by artificial intelligence on various social media and on the web to stem disinformation in view of the local elections on 3 June, the third major electoral round in the country since the legislation against the use of deepfakes was strengthened in 2023.
“With the increasing spread of AI capabilities, it happens quite often that supporters, without realizing they are breaking the law, use AI-generated support songs or posters to promote candidates’ results by inserting non-existent facts,” explains Jung Hui-hun, a computer forensics specialist at the National Election Commission (NEC).
According to experts, AI models are increasingly sophisticated and less expensive and massively fuel fake news online. In South Korea these fears are felt more strongly, because in the technologically advanced country, AI has been adopted with unprecedented speed.
“When you look at content validity, you find that it is extremely difficult to distinguish clearly, because there is such a variety of models that it is simply impossible to clearly differentiate them. Therefore, I think it is difficult to develop regulations and laws,” he adds.
The software developed by the government to detect AI offers, according to its developers, an accuracy rate of around 92%, leaving experts to examine more sophisticated content.
“If we have to take comprehensive measures against deepfakes used for election propaganda, the task is quite difficult, because freedom of expression also allows us to support candidates based on our political opinions,” says Seo Kyung-hwa, an officer of the National Election Commission’s (NEC) cyber investigation unit.
“As AI technology advances, it is increasingly difficult to distinguish it from the facts, the distinction becomes extremely complex,” he concludes.
iO Donna © ALL RIGHTS RESERVED
