For around 15 minutes on Tuesday night, blogger Perez Hilton livestreamed on TikTok with a bloody torso and a knife in hand. He seemed to be harming himself.
TikTok’s AI moderation flagged the stream within minutes, sending it to a human moderator. That person determined the video should be removed, according to a TikTok US spokesperson, but experienced technical difficulties while taking the stream down. (Wired spoke to a spokesman who called it a “moderator error,” but the company further elaborated to Vanity Fair that the moderator was hit with an “error” message when attempting to take it down.) Hilton was persistent, the spokesperson told VF, and he broadcast himself on several different streams.
Miami law enforcement received multiple calls about the incident—including one from the TikTok moderation team, the spokesperson says—and arrived at Hilton’s house to find him alone. The fire rescue team then transferred him to a local hospital to receive medical attention, per a statement from the police.
Hilton’s family posted an update on his website Wednesday afternoon. “We can confirm that Perez is receiving medical care,” they wrote, “and our family’s focus right now is on his well-being.” Born Mario Armando Lavandeira Jr., Hilton has three children but was reportedly home alone at the time.
It’s not the first graphic livestream aired on social media platforms, where users have broadcast real-time footage of mass shootings and deaths by suicide. But the 15 minutes it took to take down Hilton’s video underscores a problem that social media companies have been grappling with for years: How can you moderate something happening right now?
Livestreaming went mainstream in 2015, when Facebook rolled out its livestreaming feature. Moderation has always been a “really big challenge,” says Robyn Caplan, an assistant professor of technology policy at Duke University. In the early days, platforms often relied on users to flag inappropriate posts and human moderators to parse through potential violations. This was an incredibly intensive job, one that pushed many workers to their mental limits. In the last decade, numerous former moderators have filed lawsuits claiming they suffered trauma from spending long hours watching the worst of human behavior.
These early moderation efforts failed to stop many violent livestreams before they proliferated online. Broadcasts depicted men shooting at police and a 12-year-old taking her own life. The latter video, from 2016, was up for nearly two weeks before Facebook removed it.
