Editor’s note: This is part of a series on issues surrounding freedom of speech, focusing on “justice” in society where social media and generative artificial intelligence can quickly spread information and misinformation far and wide.
* * *
A single, scathing reply to a post by Prime Minister Sanae Takaichi on Feb. 24 ignited a firestorm on X.
Takaichi had been explaining her distribution of catalogue gifts to fellow Liberal Democratic Party members elected in the Lower House election.
The reply cut deep: “For ordinary citizens, every day is a ‘very tough’ survival with endless price hikes and heavy taxes. And yet, you privileged politician ‘sensei’ are consoling each other by exchanging luxury gifts among yourselves?”
The post quickly gained more than 158,000 views and 1,300 likes. It also drew a wave of supportive comments.
“That’s absolutely right! Their thinking is so detached from the people,” one user wrote.
Another said, “In the end, it was just a scam where they pretend to practice ‘politics for the people’ only before an election.”
The author of the reply, however, was not a disgruntled Japanese citizen.
NIGERIA ACCOUNT
An investigation by The Asahi Shimbun traced the account’s profile location to Nigeria, an African nation marked by a stark gap between rich and poor.
Tracing the account’s history, a clear pattern emerged: while it typically posted in English, it switched to Japanese specifically to reply to high-traffic posts from Japanese “expose-style” accounts that garner tens of millions of views.
Telltale signs that AI was used for its posts were left behind in the text, including fragments of the AI’s own editorial notes, such as, “Make it more inflammatory?” and, “Let’s go with something a little harsh.”
When The Asahi Shimbun sent a direct message to request an interview, the account user expressed no interest.
The Nigerian account was just one part of a much larger phenomenon.
Takaichi’s post ultimately received about 10,000 replies, and an investigation into their origins revealed another hotspot: Indonesia.
Indonesia is the world’s fourth-most populous country with an expanding economy. But like Nigeria, the Southeast Asian nation is grappling with widening inequality.
THE MAN FROM SURABAYA
One Indonesian account even accidentally posted the AI’s own instructions for the reply it had generated for Takaichi’s post: “This is an example of a reply written from the perspective of an opposition party supporter or an ordinary citizen. It includes nuances that lightly poke with sarcasm and questions to slightly provoke the other party.”
The user behind this account, a 40-year-old Indonesian man from Surabaya who identified himself as “izzi,” agreed to an interview.
He said he simply copied Takaichi’s post into X’s AI tool, Grok, with the Indonesian instruction: “Write a reply that is very blunt but also very polite.”
A real estate worker, izzi is raising a 7-year-old son and a 5-year-old daughter. He said his income became unstable after the COVID-19 pandemic and has sometimes fallen below minimum wage.
He began his Japanese-language side hustle in February to supplement his earnings.
He does not fact-check the AI-generated content and, when questioned about the impact on Japanese users, he made it clear that their reaction was irrelevant to him.
WEAPONIZING DISCONTENT
Users like izzi are known in Japan as “inpure-zonbi,” or “impression zombies.” The term spread after X launched a program in 2023 to share ad revenue with creators based on views.
With the world’s second-largest X user base, Japan became a prime target.
Initially, the replies from impression zombies were easy to spot. The posts were riddled with unnatural Japanese, and their inability to hold a conversation earned them the “zombie” label.
But the game has changed.
“With the advent of AI, their level has clearly improved,” said a Japanese language teacher who studies the phenomenon on X under the name “Masayasan.”
Having analyzed more than 100 such accounts, Masayasan believes they have mastered “rage bait”—the technique of intentionally provoking anger to maximize reactions.
“Japan, where posts frequently ‘go up in flames,’ is particularly vulnerable,” the teacher said.
In April, X suspended izzi’s account from its monetization program for sending a large volume of replies. He immediately created a new one.
He vowed to continue earning money on X for as long as he can, explaining that getting more “views” is all that matters to him, and the content itself is inconsequential.
AloJapan.com