Following the massive earthquake that struck Myanmar on Friday (Mar 28) that killed nearly 3,000 people, false videos claiming they show the aftermath of the quake have been circulating on social media. However, the video was fact-checked, proving it was generated using artificial intelligence.

Advertisment

Also read | Myanmar earthquake: Death toll nears 3,000, what are the survival chances of people trapped under rubble?

‘Inconsistency’ reveals fake video

According to an AFP report, experts pointed out the “inconsistency” of different movements in the clip, showing telltale signs that it was created with AI.

Advertisment

“Myanmar Earthquake. 17M people Affected. God save all people. Pray,” said the text in the video shared on Facebook.

Also read | Smallest coffins are the heaviest: Quake-hit Myanmar mourns death of 12 pre-school children

The false clip, which was shared a day after the 7.7 magnitude quake hit Myanmar, followed by a 6.7 magnitude aftershock, was also shared on TikTok and YouTube and gained millions of views.

Advertisment

According to AFP, in some of the videos, a TikTok username @the.360.report could be seen. While the original video was not found on the account, many of the videos shared by it had watermarks of a US-based AI generative company, Runway, and video generator platform MINIMAX Hailuo AI.

Also read | Myanmar earthquake: India hands over 50 tonnes of relief material as death toll surpasses 2,000

Shu Hu, head of Purdue University's Machine Learning and Media Forensics Lab in the United States, told AFP that he believes the video was AI-generated.

“The primary indication is that the individuals depicted in the video remain unnaturally stationary, whereas the fire visible in the distance exhibits realistic motion. This inconsistency strongly suggests artificial synthesis or manipulation,” he explained.

An AFP also confirmed that no such gaping holes in the streets were found anywhere in Myanmar following the quake, as seen in the video.

Also read | Myanmar declares week of mourning as death toll surpasses 2,000

How to identify AI-generated images and videos?

AI-generated images and videos often have inaccurate rendering of background details and depth of scale. They also have slight distortions like pattern noise, unusual shifts or flickering in the background scenery, according to Forbes.

In deepfake videos, there are chances of misalignments between lip movements and spoken words. A person in an AI-generated video may also have an awkward expression, unnatural blinking or unusual microgestures, which result in a lack of congruence between the emotional tone of the spoken words and the displayed reactions.

(With inputs from agencies)