A new example of the risks of artificial intelligence in terms of disinformation. Several verified Twitter accounts broadcast fake images likely generated by AI, which claimed to show explosions at the Pentagon and the White House on Monday, May 22.
american media Vice reports that “several dozen” verified accounts with often substantial audiences have shared footage purporting to show a plume of smoke from the explosions, which reportedly occurred near the White House and Pentagon on Monday morning.
These images were disseminated by certified Twitter accounts such as @financialjuice (386,000 subscribers) or @WarMonitors (204,000 subscribers), some posing as real media such as the “@BloombergFeed” account (since suspended), and even by the official Russian account. state media Russia Today, adds Vice.
Other accounts conveyed this rumor in writing, such as the @DeItaone account (650,000 subscribers).
Lots of inconsistent details
But many clues pointed out by Internet users suggest that these images were generated by an artificial intelligence like Midjourney. Just zoom in on the images to find many blemishes (inconsistent window bars, misshapen cars parked near the White House).
And above all, there is no other image, point of view or source on the ground that corroborates this explosion (which, however, would be a major event and visible for miles around).
The Arlington Police Department later confirmed on Twitter that “there were no explosions or incidents on or near Pentagon grounds.” Some accounts that shared these images have since removed them.
By Elon Musk’s decision, any Twitter account can get the blue badge and increase its visibility by paying the Twitter Blue subscription for $8 per month, without identity verification. A tool that has allowed a greater dissemination of these misleading images.
Source: BFM TV
