Online Platforms Like Twitter Are Missing a Brutal Wave of Hate Speech in Japan

Immediately after Shinzo Abe was assassinated by a gunman on July 8, a rumor quickly spread on Japanese social media. It falsely claimed that the suspect was a “Zainichi Korean.” The term generally applies to descendants of Korean people who emigrated to Japan between 1910 and 1945—a period when Japan occupied Korea. They are the most targeted minority in Japan and suffer from virulent online abuse.

Last summer, online hate turned into real life violence. A 22-year-old man allegedly set fire to and destroyed seven buildings in Utoro—the ethnic Korean district of Uji, a city of some 185,000 on Japan’s main island Honshu. Although there were no casualties, the attack horrified Zainichi Koreans across the country.

On August 30, he was sentenced to four years in prison.

The suspect said that the purpose of the attack was to make Koreans afraid to live in Japan. According to media reports, he was radicalized by reading anti-Korean comments that had been posted by readers of Yahoo! News Japan and one of his motives was to earn notoriety among those users.

Yahoo! News Japan reportedly has just 70 content moderators to police an estimated 10.5 million comments each month. It is the most popular news site in the country, and articles about the Utoro incident attracted a torrent of hateful comments and disinformation about Zainichi Koreans.

In a written statement to TIME, the platform said it uses “AI and [moderators] to properly eliminate some malicious users and posts” and that it has cooperated with government agencies to do so. But many of those posts remain on the site today.

Around the world, and on every continent, major tech platforms have attempted to strike a balance between allowing people to speak freely while protecting others from users who post hateful content. To do so, they’ve implemented content moderation practices that are often inadequate or particularly hard on moderators—even despite warnings from users and employees who moderate content. Japan is no exception.

But in Japan, unlike the U.S. and other countries, online problems may fester out of sight. Japan’s digital culture receives little attention from the international media and researchers beyond high-profile pop culture phenomena like anime and video games. Matt Alt, a Tokyo-based writer, says this is in part because deciphering Japanese online discourse requires a high degree of Japanese literacy. Within the country, what happens online tends to stay online, says Alt: “There is sort of a barrier between online happenings and mass media in Japan. More so than the West.”

Japanese far-right netizens called “netto-uyoku” flock to Yahoo! News Japan and other platforms like Twitter and Japanese Wikipedia that allow anonymity. They use the sites to spread historical revisionism and stoke xenophobic views of Korea and China.

Twitter has over 45 million Japanese users, making Japan its second largest market. It has a policy that “prohibits statements of exclusion based on race or ethnicity,” according to a Twitter spokesperson, who specifically added that “statements of exclusion or violence towards the Zainichi Koreans will be subject to enforcement.”

But Zainichi Koreans are frequent targets of abuse on the site, where they are derided as “cockroaches,” “cancer,” “illegal immigrants,” and “chon” (a highly derogatory term), while being told to “go back to your country.” The last attack is particularly painful, given that the ancestors of many Zainichi Koreans were forcibly sent as laborers to Japan during the colonial era. One Zainichi Korean described it to TIME as “the murder of the soul.”

Meanwhile, with one billion monthly page views, Japanese Wikipedia is the most visited edition of Wikipedia after English. It has played a crucial role in whitewashing war crimes committed by Imperial Japan in China and Korea, some experts say. The Zainichi Korean page contains many misleading claims and reinforces a stereotype of Zainichi Koreans as criminals. One of Wikipedia’s volunteer editors said in an email to TIME that “Japanese Wikipedia has been hijacked by netto-uyoku.”

The Wikimedia Foundation, which operates Wikipedia, dismisses the claim. It said in a statement to TIME that it had investigated historical revisionism on Japanese Wikipedia and found “some presence of right-wing users who have possibly attempted to control the content on certain pages” but the abuse didn’t seem frequent or sufficient enough to enforce a ban.

The foundation added that it is not involved in decisions about content on the site, as those decisions are made by volunteer editors.

The foundation also added that its “volunteers have included more relevant, verified historical context in [disputed] articles,” although much disinformation remains.