Two leaders of a fledgling Japanese political party unveil a red logo reminiscent of Chinese Communist Party imagery. A candidate campaigns in freezing weather wearing a tank top. Grannies vent in public about what they see as political nonsense. Or do they?
Whatever the outcome of Sunday’s Lower House election, the race will be remembered not least for having witnessed the full-blown debut of “AI slop” in Japanese politics.
As more voters in Japan turn to social media for political updates and policy debates, the internet is increasingly looking like a dumping site for fake and dubious news and content.
Behind the deluge of such content is the ease with which generative AI can create authentic-looking videos and images — and the opportunities for anyone to make quick money, experts say.
Major social media platforms, such as YouTube and TikTok, offer financial rewards to channels with high numbers of views.
“The mechanism through which fake news and content of uncertain authenticity spreads hasn’t changed much since the previous election,” said Shinichi Yamaguchi, a professor at the International University of Japan in Niigata Prefecture. “But what is decisively different in this election is how generative AI technology has advanced.”
One fake image appeared shortly before the election, when the newly-formed Centrist Reform Alliance (CRA) revealed its logo. The image featured a fake red logo instead, with users suggesting links to the Chinese Communist Party.
In another example, one YouTube user has uploaded over 120 videos featuring elderly women who make rapid-fire rants about political developments, with titles such as “An AI granny moved by Takaichi Cabinet members visiting the Ise Shrine.” The channel has racked up nearly 6 million views.
Some AI-generated content, like the AI granny series, is easy to spot as fake and is marked as such. But other pieces are convincing enough that anyone scrolling through social media could easily be misled.
The X platform uses its proprietary AI chatbot Grok to correct and supplement information, but its service is not foolproof.
One video posted by a CRA candidate showed a huge turnout at his campaign rally in a Fukuoka park. When the official CRA supporter account shared the video on X, the post was swamped with comments from people questioning the video’s authenticity.
One user asked Grok to weigh in. The AI chatbot concluded that the video was likely AI-generated, despite the candidate’s assertion that it was real.
The ever-increasing volume of information on the internet makes anyone vulnerable to misinformation and disinformation, Yamaguchi says. A 2024 report released by his team included a survey of 3,700 people across generations, from teens to those in their 60s. It found that 51.5% of people had…