‘Verified’ Twitter Accounts Spread Pentagon Explosion Hoax


Breaking news: The U.S. Pentagon is NOT under attack. But that didn’t stop false reports of an explosion from circulating widely online along with a seemingly AI-generated image of billowing black smoke near the Department of Defense headquarters on Monday.

Multiple blue-check “verified” Twitter users re-posted the fake photo along with text insinuating an active emergency situation was ongoing at the Pentagon, and later the White House. Accounts amplifying the false narratives included at least one that appeared to be actively impersonating Bloomberg News (the offending account, @BloombergFeed, was later suspended).

Russian state media outlet RT news further gave the hoax story a boost when it posted about the fake incident on its gold check-wielding, paid, verified account. Though RT eventually deleted its tweet hours after initially posting it, more than 38,000 people viewed the post. Republic TV, a right-wing, English-language news channel in India also broadcast an entire news report based on the fake image.

The Pentagon became a trending topic on Twitter, and remains on the trending list, as of publication. In a search for “Pentagon” on the platform, most of the top posts appear to be debunks of the hoax explosion, while a few continued to promote the fake news. Far fewer discussed the recently released astronomically high Pentagon budget, and a new investigative CBS News report suggesting that contractors are overcharging the government.

Things got out of hand enough that Virginia authorities had to step in to reassure that there had not actually been any explosion or other incident at the Arlington DOD headquarters. The Pentagon Force Protection Agency and Arlington Fire & EMS and posted a statement on Twitter refuting the hoax. “There is NO explosion or incident taking place at or near the Pentagon reservation, and there is no immediate danger or hazards to the public.”

But beyond Twitter, the likely AI-generated explosion scare had real-world reverberations as well. Minutes after one of the first posts alleging an incident at the Pentagon emerged online, Insider reported that the U.S. stock market briefly dipped—though stock values quickly recovered.

What may not recover quite as quickly is Twitter’s reliability and authority as a place to share and find news. Since Elon Musk overhauled his social media platform’s verification system last year, and moved to eliminate most legacy checks last month, the only accounts with checkmarks remaining on the platform are those who pay for Twitter Blue or Twitter’s Verified Organization program and a small number of institutions and government agencies granted free, gray checks.

Previously, Twitter verification was only available to accounts that were “authentic, notable, and active.” Users generally had to apply for a blue check, and most bids were rejected. But that verification process doesn’t exist anymore. Instead, for $8 per month and without much (if any) oversight, Twitter Blue can access an aura of authority on the site.

Basically as soon as pay-to-play checkmarks became available, imposter accounts for celebrities and major companies became a problem. Though Musk has attempted to institute a fix, today’s incident demonstrates that the issue persists.

Compounding the problem of online impersonation is, of course, the rapidly improving generative AI technology that allows basically anyone to instantly produce convincing-enough images, video, or sound clips of anything they choose to. Using AI-generated text and images, content farms have popped up across the internet, spewing completely fabricated, yet clickable content. Previously, fake “photos” (faux-tos?) of Donald Trump’s arrest made big waves online, leading some experts to call for regulation or mandatory watermarks on generated images.

Though the fake Pentagon explosion image hasn’t been verified as AI-created, it almost certainly was. Inconsistencies in the fence and building windows seem very much in line with the type of hallucinations AI frequently makes.

Source link


Please enter your comment!
Please enter your name here