As Hurricane Melissa, a massive Category 5 storm, bore down on Jamaica on Monday, a deluge of artificial intelligence-generated videos and images flooded social media platforms, creating a hazardous information environment. Dozens of fabricated videos, many bearing the watermark of OpenAI’s text-to-video model Sora, circulated widely, threatening to overshadow official safety alerts and critical updates from authorities. This wave of synthetic media highlights a dangerous new front in the battle against misinformation during natural disasters, where realistic fakes can easily mislead the public and dilute the urgency of emergency communications.
The proliferation of these AI-generated fakes prompted immediate concern from government officials and cybersecurity experts, who warned that the content could lead to catastrophic consequences. The fabricated media ranged from the absurd to the deeply misleading, featuring everything from locals partying in defiance of the storm to emotionally charged scenes of fabricated suffering. Experts note that the ease with which these convincing fakes can now be created and distributed has created an information paradox: as more content becomes available, the public may become less informed and less prepared for real-world threats.
Anatomy of the Deception
The AI-generated content that spread across platforms like TikTok and WhatsApp depicted a wide array of fabricated scenarios designed to capture attention and evoke strong emotional responses. Some videos showed dramatic, entirely fictional newscasts detailing the hurricane’s approach, while others featured realistic-looking but fake scenes of severe flooding. A recurring and long-debunked trope, the appearance of sharks in flooded urban areas, was also resurrected through AI generation.
Beyond sensationalist imagery, another category of fakes aimed to downplay the storm’s severity. These videos showed people appearing to be locals—some voiced with strong Jamaican accents that seemed to reinforce stereotypes—boating, jet skiing, and otherwise ignoring the danger of what forecasters warned could be the island’s most violent storm on record. Other clips were designed to manipulate empathy, including a fabricated video of a woman holding babies under a roofless home, which garnered numerous prayers and concerned comments from viewers who believed it was real.
Officials Scramble to Counter Fakes
In response to the flood of misinformation, Jamaican officials moved to reassert control over the narrative and guide citizens toward reliable sources. Senator Dana Morris Dixon, the nation’s information minister, participated in a press briefing specifically to provide “correct information” about the approaching storm. “I am in so many WhatsApp groups, and I see all of these videos coming. Many of them are fake,” Dixon stated. She urged the public to “please listen to the official channels” to stay informed about the hurricane.
Despite these efforts, the AI-generated videos continued to gain traction. An AFP review of comment sections on platforms like TikTok found that many viewers were unaware the images were synthetic, even when they included a watermark from the AI model that created them. For example, one popular video showed an elderly man defiantly yelling at the hurricane that he would not “move for a little breeze,” prompting commenters to offer prayers for his safety and ask for updates on his property, demonstrating the real-world impact of the fabricated content.
The Accelerating Threat of Synthetic Media
Expert Warnings on Public Safety
Experts in meteorology and cybersecurity voiced grave concerns about the potential for such misinformation to cause direct harm. Amy McGovern, a professor at the University of Oklahoma whose research involves using AI to improve weather forecasting, emphasized the danger of undermining official alerts. “This storm is a huge storm that will likely cause catastrophic damage, and fake content undermines the seriousness of the message from the government to be prepared,” she told AFP. McGovern warned that the continued spread of such content will eventually “lead to loss of life and property.”
The Technology Fueling the Fakes
Hany Farid, a professor at the University of California, Berkeley and co-founder of GetReal Security, noted that the hurricane-related fakes underscore how advanced text-to-video models have “accelerated the spread of convincing fakes.” These powerful tools allow users to generate hyper-realistic clips with ease, overwhelming the information ecosystem. The clips identified by AFP spread primarily on TikTok, where the platform’s policy requiring users to disclose realistic AI-generated content was only inconsistently applied. Farid described the situation as “the paradox of the information age,” where “we are becoming less informed as a public as the amount of information increases.”
A Disturbing New Norm for Disasters
The events surrounding Hurricane Melissa are not an isolated incident but part of a growing trend of AI-generated misinformation during natural disasters. In 2024, a similar influx of fake images spread online following Hurricanes Helene and Milton. These events force the public and journalists into a difficult position, having to second-guess the authenticity of every dramatic image, such as a dumpster sitting atop a house, which in one case was real but was initially met with skepticism.
Social media platforms have struggled to keep pace with the evolving threat. Some have experimented with heavy-handed approaches, such as Instagram temporarily making the term “FEMA” unsearchable to curb the spread of misinformation. However, such measures are often reactive and incomplete. As AI technology becomes more sophisticated and accessible, the challenge of separating fact from fiction during fast-moving emergency situations will only intensify, making media literacy and a reliance on verified sources more critical than ever.
