NEW YORK — No, Katy Perry and Rihanna didn't attend this year's Met Gala. But that didn't stop the AI-generated images from fooling some fans into thinking the stars were on stage at fashion's biggest night.
A deepfake image depicting several celebrities attending the Metropolitan Museum of Art's annual fundraiser went viral online between Monday and early Tuesday.
Some astute social media users spotted the discrepancy, and the platforms themselves, such as X's Community Notes, were also quick to point out that the image may have been created using artificial intelligence. For example, one clue that the viral photo of Perry in a flower-covered gown was fake was that the staircase carpet matched that of the 2018 event, and this year it was lined with live leaves. That means it wasn't a greenish fabric.
Still, some were fooled, including Perry's own mother. Hours after at least two AI-generated images of the singer began circulating online, Perry reposted them to her own Instagram, confirming what she thought were her real Met Gala appearances. She has attached a screenshot of a text that appears to be from her mother praising her.
“Lol, mom, you got an AI too, be careful!” Perry responded to the exchange.
A representative for Mr. Perry did not immediately respond to a request from The Associated Press for further comment and information about why Mr. Perry did not attend Monday night's event. But in the caption of his own Instagram post, Perry wrote, “I couldn't go to the Met, so I had to work.” The post also included a video of her singing silently.
Meanwhile, a fake image of Rihanna wearing a stunning white dress embroidered with flowers, birds and branches also surfaced online. The multihyphenate was originally confirmed as a guest at this year's Met Gala, but Vogue officials said before closing the carpet on Monday night that she would not be attending.
People magazine reported that Rihanna was suffering from the flu, but a representative did not immediately confirm the reason for her absence. A representative for Rihanna also did not immediately respond to a request for comment on the AI-generated image of Rihanna.
Although it is difficult to determine the source of these images, the realistic Met Gala backgrounds seen in many images suggest that the AI ​​tool used to create the images was trained on several images from past events. This suggests that there is a high possibility that the
Getty Images, the Met Gala's official photographer, declined to comment Tuesday.
Last year, Getty sued Stability AI, a leading London-based AI image generation company, for allegedly copying more than 12 million photos from Getty's stock photo collection without permission. Getty later launched its own AI image generator trained on his company's work, but attempts to generate what it called “questionable content” were blocked.
This isn't the first time we've seen generative AI, a type of AI that can create new things, being used to create fake content. Deepfakes of images, videos, and audio of celebrities, from Pope Francis to Taylor Swift, have garnered a lot of attention online in the past.
Experts say each case highlights growing concerns about the potential for misuse of this technology, particularly to carry out disinformation and fraud, identity theft and propaganda, and even election manipulation. .
Casey Myers, professor and director of graduate studies at Virginia Tech's School of Communication, pointed to the impact of Monday's AI-generated Perry image, saying, “Once upon a time, to see was to believe. And now, to see is to believe. It's not about believing.” She said: “[If]even the mother can be fooled into thinking the image is real, that shows how sophisticated this technology is.”
Using AI to generate images of celebrities in pretend extravagant dresses (which can easily be proven to be fake at highly publicized events like the Met Gala) seems relatively harmless. Although there may be more serious or harmful uses of this type of technology, Myers et al.
For example, earlier this year, X (formerly Twitter) temporarily blocked some searches after explicit and abusive fake images of Swift began circulating online. Of course, celebrities are not the only victims of non-consensual deepfakes, and advocates highlight particular concern for victims who have few protections. Research shows that explicit AI-generated content disproportionately harms women and children, including the alarming case of her AI-generated nudes being distributed to high schools. Masu.
And in what is an election year for several countries around the world, experts also continue to point out the potential geopolitical implications of deceptive AI-generated content.
“The implications here go far beyond personal safety, and really touch on things like national security, the safety of society as a whole,” said David, an associate professor at George Washington University and principal investigator on the study.・Mr. Broniatowski said. Institute for Trustworthy AI in School Law and Society.
Leveraging the capabilities that generative AI offers while building an infrastructure that protects consumers is a challenge, especially as the commercialization of this technology continues to grow at an extremely rapid rate. Experts point to the need for corporate accountability, universal industry standards, and effective government regulation.
As governments around the world try to catch up, technology companies are taking the lead when it comes to managing AI and its risks. Still, significant progress has been made in the last year. The European Union reached an agreement in December on the world's first comprehensive AI regulation, but the law will not come into force until two years after final approval.
_____________
Associated Press writers Matt O'Brien in Providence, Rhode Island, and Kelvin Chan in London contributed to this report.