AI-generated videos of Hurricane Melissa flood social media • FRANCE 24 English

Key Concepts

AI-generated content: Images and videos created by artificial intelligence.
Deepfakes: Realistic but fabricated media, often used for misinformation.
AI detection tools: Software used to identify AI-generated content.
Watermarks: Visible or hidden indicators of AI generation, often from specific tools like Sora or VEO.
Misinformation/Disinformation: False or misleading information, spread intentionally or unintentionally.
Hurricane Melissa: A recent hurricane that served as a backdrop for the spread of fake content.
Official warnings: Alerts issued by authorities to ensure public safety.
Social media platforms: Online services like TikTok and X where fake content can spread rapidly.

Fake Content During Hurricane Melissa

The video discusses the proliferation of fake images and videos on online platforms like TikTok and X in the wake of Hurricane Melissa. These AI-generated materials, ranging from fake news reports to realistic depictions of sharks swimming in flooded streets, prompted official warnings from Jamaican authorities.

Exhibit A: Fake Photo of Damage

Description: A photo purporting to show damage from Hurricane Melissa in Black River, Jamaica.
Evidence of Fakeness: The timestamp on the photo indicated it was posted four hours before the storm hit (1 p.m. on October 28th).
AI Detection: Fact-checkers at Lead Stories used the Hive moderation tool and found the image to be 99.8% likely AI-generated.

Fake Video of the Storm's Eye

Description: A video claiming to show the eye of Hurricane Melissa from 30,000 feet above.
Evidence of Fakeness: A community note on X pointed to the American Weather Agency's website, which explains that cyclones in the Northern Hemisphere move counterclockwise, while this video depicted a clockwise rotation.
Comparison: A real video taken by a US Air Force Reserve crew (hurricane hunters) who flew into the storm's eye for data collection was shown for comparison, highlighting the visual similarity between real and fake content.

Shark Videos and AI Watermarks

Description: Alarming videos showing sharks apparently swimming in flooded Jamaican streets or swimming pools due to storm surge.
Impact: These videos garnered hundreds of thousands of views on TikTok and appeared in top search results for Hurricane Melissa.
Telltale Sign: A closer look at these videos revealed a blurry, hidden mark indicating AI generation.
Specific AI Tools: These watermarks are often from OpenAI's Sora or Google's VEO text-to-video tools.
Example: A humorous video showing men smoking with dolphins in the middle of a storm clearly displayed the "Sora" watermark.
Masking Attempts: Some creators attempt to hide these watermarks, as seen in a fake news clip of a reporter in Kingston, where the Sora mark was obscured by a "like and follow" graphic.

Consequences of Fake Content

Public Deception: Comments on these posts indicated that people were believing the fake content.
Official Warnings: Jamaican authorities issued warnings for people to rely only on official information.
Underestimation of Danger: Some AI-generated videos caused people to underestimate the storm's danger and take unnecessary risks.
Drowning Out Safety Messages: Viral fake content clutters social media and hinders the spread of real safety information.
Platform Response: TikTok removed some fake videos after a report by AFP, but many remained available across social media platforms.

Conclusion

The spread of AI-generated fake content during Hurricane Melissa highlights a significant challenge in distinguishing reality from fabrication online. While some of this content may seem harmless or even humorous, it can have serious consequences, leading to public misinformation, underestimation of risks, and the suppression of vital safety messages. The presence of AI watermarks, though sometimes hidden, remains a key indicator for identifying such content, and official sources are crucial for accurate information during emergencies.