The 9 Dumbest AI Pranks People Fell for in 2023

Bill Gates, who has become something of a boogeyman to the antivax movement in the Covid-19 era because of his philanthropic foundation’s work on vaccine research and distribution, was the target of this deepfake video that first surfaced on TikTok in March. The clip appears to show Gates being grilled in a TV interview by anchor Sarah Ferguson of Australian Broadcasting Corporation News — first about whether he really created the Windows operating program, then as to how Covid shots are allegedly causing “injuries, side effects and death.” Gates protests somewhat before finally announcing, “Okay, that’s enough, this interview is over.” The real interview, which aired in January, includes none of this dialogue, nor a dramatic exit. But for antivaxxers, it hardly mattered either way. “Not sure if this is real but the content is true,” wrote one Instagram user who shared the AI-altered footage.

An image that appeared to show a thick plume of black smoke in the aftermath of a supposed explosion near the Pentagon gained traction in May, and was spread by X (then Twitter) accounts with paid verification check marks, including the Russian media outlet RT and a two others pretending to be affiliated with Bloomberg News. This was just weeks after site owner Elon Musk had done away with traditional verification checks for journalists and media organizations — increasing confusion over which sources could be trusted. The stock market took a brief hit in the ensuing panic, with a bump in gold prices from investors who may have been worried about a destabilizing disaster. Federal authorities and local emergency personnel in Arlington, Virginia, soon confirmed that no explosion had taken place, with fact-checkers concluding that the picture had been created with an AI program.

Receiving a phone call or voice message from a loved one in distress is an understandably scary. Alas, scammers can take advantage of that fear and alarm by using AI tools to “clone” a person’s voice — all they need is a few seconds of video from social media — and applying it in an extortion scheme. Parents have heard what sounded like their own children begging to be rescued from kidnappers or saying they’re in trouble with the police and need bail money. Such scams are on the rise, with a survey from the computer security company McAfee finding that a quarter of adults have experienced one, with many duped out of thousands of dollars. There are, of course, ways to confirm that it’s really your relative that you’re talking to — a handy trick is to hang up and immediately call back.

This was it: the “photo” that seemed to wake everyone up about how AI-generated art can fool the masses. Or was it that we all just badly wanted to believe that Pope Francis has drip? Either way, the Midjourney creation depicting the pontiff in a Balenciaga-inspired white puffy jacket, first posted on Reddit, went mega-viral at the end of March and triggered a slew of articles warning us not to believe everything we see online. Among those who took the image at face value (apart, perhaps, from fans of The Young Pope) was Chrissy Teigen, who tweeted that she believed the coat was real and “didn’t give it a second thought,” joking that she won’t survive “the future of technology.” Then again, who will?

As Rolling Stone‘s Ej Dickson reported in October, women who date dopey men are at higher risk than ever of losing a boyfriend’s affections to AI sex bots that have infested Instagram and other social platforms. Some of these fake personalities are used to sell adult content, though others have seen unsuspecting and besotted chumps fall prey to credit card fraud and blackmail. It goes to show that the average straight male doesn’t have especially high standards when it comes to proper grammar and syntax in a sexting partner — hell, the person doesn’t even have to exist. A little reverse image searching when you get a random explicit DM from a barely-clothed model doesn’t hurt, fellas. (Aside from a slight bruising of the ego, anyway.)

Brands faced a barrage of hate from far-right extremists this year over promotions that involved or catered to the LGBTQ community, and Target was trashed almost as often as Bud Light. At issue was the chain’s Pride Collection of rainbow-festooned clothes and merch, a selection of products that led conservatives to call for a boycott on the absurd pretext that Target was “grooming” kids into becoming gay or trans. (The company ended up pulling some items after patrons threatened employees.) In an apparent effort to heighten this outrage, someone disseminated AI-generated images of kids wearing shirts with pentagrams and store displays that included Baphomet-like goat heads. These made the rounds in conservative religious corners of social media, with users falsely claiming Target was selling “Satanic” apparel. Never let the truth get in the way of a good smear campaign!

Sometimes a deepfake is so nice you have to circulate it twice. Such was the case with a bogus video of President Biden appearing to announce a U.S. draft shortly after the militant group Hamas launched an attack Israel in October, leading Israel to extensively bomb the Gaza Strip in retaliation. Originally shared in February by conservative publication The Post Millennial with the caption “AI imagines what would happen if Biden declares and activates the Selective Service Act and begins drafting [20-year-olds] to war,” the clip sees the fake president declare, “Remember, you’re not sending your sons and daughters to war. You’re sending them to freedom.” When it resurfaced on TikTok amid the Middle East war later in the year, the White House had to deny for the second time that any such speech had taken place. Which probably won’t stop misinformation peddlers from using it again. 

OpenAI’s text bot ChatGPT was one of the most-discussed AI tools in 2023. Millions experimented with the software, and not always in harmless ways: lawyers relying on ChatGPT to write a court filing, for example, wound up accidentally citing fictitious cases. On the flip side, college professors overzealously checking homework for evidence of ChatGPT assistance sometimes falsely accused students of cheating. Still, one of the worst consequences of unleashing this technology is among the most banal: grifters have used the program to spawn hordes of social media bots that hawk cryptocurrency to gullible human users. Indiana University Bloomington researchers uncovered such a network in May on X, warning it was likely just the tip of the iceberg — this despite site owner Musk’s occasional claims about eradicating bots and spam from the platform.

Much has been made of the connections that the late pedophile and sex trafficker Jeffrey Epstein had to the ultra-wealthy elite in his lifetime, and it’s no secret that he and Donald Trump were especially close for years, beginning in the late 1980s. (The two later had a falling out.) A video of them partying together at Mar-a-Lago in 1992 frequently circulates as evidence of their intimacy. It seems, however, that this was not damning enough for the presumable Trump critic who employed an AI image model to generate a picture of them cozying up with young girls on a couch. The fake picture popped up whenever a blue-wave Democrat wanted to remind followers of the former president’s Epstein ties — an odd strategy given the preponderance of real photos and footage showing them together. The counterfeit doesn’t stand up to scrutiny, either: we’re pretty sure Epstein had legs.