Strong preventive measures required for AI-sourced disinformation

Ahead of major elections in Korea and overseas, including the Korean general elections and the U.S. presidential election next year, there are growing concerns about the possibility of widespread disinformation using artificial intelligence technology. As 'deep fake' technology, which synthesizes and manipulates videos and audio, is rapidly evolving in tune with the development of AI, there will be more attempts to exploit it in elections. The political threat of AI, which was not an issue until the 2020 U.S. presidential election, has set off an alarm in the political sphere.

Generative AI, which is evolving day by day, has reached a point where even non-experts can easily create fake content using commercially available programs. When The Dong-A Ilbo tried this with an expert, it took only 10 seconds to create a false statement in the voice of a broadcast news anchor and one minute to change the background image of a video. The problem is that technological limitations are being quickly overcome with new versions, making it extremely difficult to tell what's real and what's not.

If used in elections, it can be manipulated to produce false campaign speeches or interviews of candidates in a matter of seconds. There are also ways to manipulate a rival candidate's voice to send false messages via the automatic response system and create fake news announcing a candidate's resignation close to the end of campaigning. Celebrities can also be faked as if they are endorsing a certain candidate and thus used in political ads and fundraisers. This is already a problem in foreign countries, including the U.S. Just as we recently saw the doctored photos of the Pentagon attack, we can expect to see more such attempts to sway votes by stoking security fears.

South Korea will be no exception when the general elections take place in April next year. We are concerned about the possibility of AI being manipulated by those who want to spread inflammatory disinformation to take advantage of social conflicts that divide the country between liberals and conservatives, the elderly and the young, genders, and the rich and the poor. Politics in Korea is already suffering from the ill effects of rigid fandom politics awash with sensationalized claims and accusations. Content spreads faster on social media in Korea than in any other country. There is a risk that disinformation tailored to target various voters could sway public opinion and distort election results, leading to consequences jeopardizing the very foundation of democracy.

Experts say that to address these challenges properly, we must start preparing countermeasures at least a year and a half before election day. The European Union is pushing to legislate an AI regulation law to prevent adverse effects, such as obliging the creators to insert the label of ‘AI-generated content.’ As it can be already too late, Korea should refer to overseas movements and rapidly prepare institutional measures to prevent political misuse of AI and to raise public awareness. Those who intentionally manipulate or attempt to influence elections by going beyond freedom of expression‎ should be held sternly accountable.