AI fakes, graphic footage deepen Nepal flood trauma online

AI slop and ragebaiting graphic content are raising alarms about misinformation and secondary harm for both victims and viewers of online content proliferating in the aftermath of the Nepal floods.

A graphic video showing two men stealing earrings from the bodies of victims of the Nepal floods had garnered millions of views on YouTube as of Monday. The footage showed the back of a victim's dead body without any blurring of the image. Another video, showing a body with severe injuries to its lower half being recovered, had been shared tens of thousands of times.

AI-generated fabrications and footage from previous disasters falsely presented as scenes from Nepal were also widespread.

One paid account on X claimed that a tourist filming floodwaters in Nepal was suddenly swept away, but the footage was actually filmed along the Qiantang River in Zhejiang, China, in 2024.

Another video purporting to show an elephant rescuing a man being swept away by floodwaters surpassed 1 million views within a day of being posted, but was found to have been filmed in India during a previous incident.

The spread of such content can be attributed in part to social media algorithms that favor disaster footage because it tends to drive high levels of engagement.

“I’m aware that false videos and information have recently been spreading on social media,” said a trekking guide who has worked in Nepal for 12 years. “Trust and truth are more important than anything during a disaster, and it is deeply concerning to see actions that undermine the morale of people working together to help.”

Internet users have also voiced concern. “Bodies are shown uncensored in short-form videos,” said one user in the comment section of an online community post.

“My heart starts racing before I even know what’s on the screen, so I swipe to another video,” said another.

“The posts showing what was supposedly happening in Nepal were so shocking that I shared one, but I took it down after finding out it was AI-generated,” an office worker surnamed Seong said. Some Korean media outlets also took down articles that used AI-generated content for their stories.

Nepalese authorities formally asked Meta and TikTok on Friday to swiftly remove AI-generated videos and false posts and reduce the visibility of content that portrays the disaster in an excessively sensational manner.

This is not the first time indiscriminately circulated photos and videos have caused harm in the aftermath of a disaster. During the 2022 Itaewon crowd crush, videos showing victims with cyanosis — a bluish discoloration of the skin caused by insufficient oxygen — circulated on social media, such as YouTube and Instagram, which contributed to broader social trauma.

Posting and sharing unfiltered photos, videos and information from disaster scenes on social media, as well as viewing such content, were associated with post-traumatic stress disorder (PTSD), according to a 2024 study by Yoo Woo-hyun, a professor in media communication at Incheon National University.

Similar findings have been reported overseas as well. A study of 516 people who did not directly experience the 2021 floods in Henan province, China, found that greater exposure to disaster-related short-form videos was positively correlated with higher scores for depression, anxiety and PTSD.

Experts describe the phenomenon in which people experience trauma after repeated exposure to videos or news coverage despite not being directly involved in a disaster as vicarious, or secondhand, trauma.

Starting in January, businesses are required to label AI-generated content, but individuals are not subject to the requirement under Korea's Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust.

This leaves much of the responsibility for policing such content to the platforms themselves.

“False and manipulated information during a disaster can delay rescue operations and deepen the suffering of bereaved families,” said Moon Hyun-cheol, vice president of the Korean Association for Disastronomy. “Those who maliciously spread false rumors need to face strong legal consequences.”

“As advances in AI make it easier to create false content, we’re seeing people produce and consume it almost as a form of entertainment,” said Yoo Hyun-jae, a professor at Sogang University's College of Communication. “Self-regulation by platforms has its limits, so reporting systems and institutional safeguards need to be strengthened, along with tougher penalties for those who maliciously spread such content to create a stronger deterrent.”

BY LEE AH-MI, NOH YU-RIM [[email protected]]