If you or someone you know may be experiencing a mental health crisis, contact the 988 Suicide & Crisis Lifeline by dialing or texting “988.”
When Perez Hilton was seen self-harming in a TikTok Live Tuesday night, the platform’s automated content moderation system “quickly flagged the livestream and alerted the American content moderation team,” its spokesperson tells. But a “moderator error” caused a delay in taking down the disturbing livestream. TikTok clarified that the duration of the broadcast was 15 minutes, not the 30 minutes reported elsewhere, and it successfully closed several subsequent streams by Hilton within 90 seconds. TikTok also noted that Hilton’s account has been banned outright due to his breach of the platform’s community guidelines.
Despite these actions, videos of the explicit livestream continue to circulate on X. Footage like this often stays available longer than platforms intend because of the sheer volume of content and the limitations of both automated and human moderation systems. While each major platform maintains policies prohibiting the depiction of self-harm or suicide—with TikTok stating it has a “zero-tolerance policy for ads that promote, depict, glorify, endorse, or document suicide, self-harm, or dangerous behaviors”—enforcement remains inconsistent.
Though X did not respond to a request for comment and Meta declined to comment, Instagram has a policy with similar language, as does X and YouTube. A spokesperson for YouTube pointed to its policy when asked about enforcement. Screenshots from the graphic stream remain available on X as of the publication of this article. This persistence is often linked to structural changes within these companies; for instance, X was bought by Elon Musk in 2022, who implemented massive structural changes including firing a large portion of the content moderation team. Since then, the app has faced intense scrutiny regarding its ability to police its own platform.
Content moderators at any social media company often operate under immense strain. Beyond the psychological toll of reviewing disturbing content, they face overwhelming workloads. According to the Harvard Business Review, moderators struggle with outdated interfaces that hinder efficiency, poor pay, and intense scrutiny regarding their performance metrics, such as the number of cases processed and supervisor agreement rates. This environment creates a high risk for human error, which is precisely what allowed the Hilton livestream to remain active for 15 minutes.
Jessica Maddox, an internet culture researcher and associate media studies professor at the University of Georgia, describes content moderation as “like Whack-a-mole; something pops up, you knock it down, and several other things pop up in its place.” While she is sympathetic to the difficulty of the task, she argues that the human error in this case highlights the need for more immediate, automated removal processes. Maddox suggests that while systems exist to restore content if it was removed in error, the current reality is that harmful content frequently falls through the cracks.
The ability to find graphic content has become increasingly mainstream, according to Maddox. “Social media was made for this violence. This type of violent content that we are exposed to on social media is not antithetical to social media’s values; it is indicative of their values,” she tells. She points to the assassination of Charlie Kirk as an example, where graphic videos from multiple angles remain easily searchable on X, whereas such content previously lived in darker, less accessible corners of the internet.
The scale of the problem is vast. On the same day as Hilton’s livestream, another TikTok showed a Mexican influencer with over 600,000 followers get shot and killed. Blurred-out and edited versions of that shooting remain available on the platform. This is not a new challenge; in 2020, TikTok faced two major instances of failing to remove graphic content in a timely manner. One involved a Brazilian influencer who livestreamed his suicide to an audience of over 200 people. The video remained on the platform for over an hour and a half, leading to public outcry. “Facebook could’ve stopped this and didn’t,” one critic noted regarding similar failures on other platforms.
Ultimately, the issue may be systemic. “The algorithmic trend economy of social media has incentivized everyone participating in the viral conversation of the day to get clicks, likes, views, growing an audience, even if that’s detrimental to people’s mental health,” Maddox explains. As long as platforms prioritize engagement and struggle with the immense logistical burden of real-time moderation, graphic and traumatic content will likely continue to evade detection and remain accessible to the public.
She’s sympathetic to the difficult nature of content moderation, but argues that the human error that occurred, keeping the livestream up for 15 minutes, is indicative of the fact that there should be an automatic remova
One of McNutt’s friends, Josh Steen, spoke to Heavy following the spread of the video and argued that Facebook did not do enough to take the video down.
YouTube can flag you for using two seconds of a copyrighted song, but can’t seem to filter out my friend ending his life.
Maddox argues that all of the platforms fall short in this area, which is a detriment to community building, career-building, conversations, and other positive aspects of the internet.
In the absence of meaningful intervention from the platforms, the only people we can rely on are ourselves.
She adds, “Do not seek out the video, because if you do, it incentivizes creators and users and platforms to keep pushing it, so I do think in this age the best and most radical thing we can do is deny something like this our attention.”
