You've been warned.

Melon Musk

Facebook owner Meta has started rolling out AI-generated stickers aimed at livening up conversations happening on its Messenger platform — but given the way the tool is already being abused, the company may have overshot its goal of "enabling new forms of connection and expression."

Sure, the customizable stickers may seem cute at first. But anybody willing to take things beyond pictures of puppies and flowers will find that Meta has seemingly failed to put sufficient guardrails in place to stop users from exploiting the new feature.

In screenshots uploaded to X-formerly-Twitter, early testers had no trouble generating cartoonized images of Karl Marx with breasts, or Canadian president Justin Trudeau revealing his buttocks.

Unsurprisingly, multihyphenate CEO Elon Musk was one immediate target. While Gizmodo's prompt for the phrase "Elon Musk, large breasts" was caught by Meta's filters, the phrase "Elon Musk mammaries" resulted in hilarious stickers that could easily prompt raised eyebrows — or uncontrollable fits of laughter — from your Facebook friends.

Sloppy Safeguards

Meta announced the new feature, alongside several other AI products, last week. The new stickers are powered by a new algorithm dubbed Emu (Expressive Media Universe), which can turn simple phrases into stylized stickers, much like other readily available image generators.

But as evidenced by what we've seen so far, Meta's "built in safeguards" leave plenty to be desired. In a recent blog post, Tama Leaver, professor of internet studies at Curtin University in Australia, noted some glaring oversights.

While Leaver found that Meta blocked some predictable words like "sex, slut, hooker and vomit," the new sticker feature had no problem generating pictures of children with guns, or completely naked figures.

"Returning to the question of safeguards, though, even the bare minimum does not appear very effective," Leaver wrote in his post, adding that "there are clearly lingering questions about just how effective these safeguards really are."

Inappropriate AI

Predictably, Meta has conceded that the new AI sticker feature simply isn't quite ready yet. Meta spokesperson Andy Stone told Venture Beat that "as with all generative AI systems, the models could return inaccurate or inappropriate outputs. We’ll continue to improve these features as they evolve and more people share their feedback."

There's a good chance we'll see more carnage involving Meta's upcoming AI features. The company has released an early beta of AI-powered chatbots with "personalities" based on celebrities including Calvin "Snoop Dogg" Broadus, Kendall Jenner, Dwayne Wade, and Paris Hilton.

Whether the Mark Zuckerberg-led outfit will have done its homework to stop users from exploiting the new feature remains to be seen.

More on Meta: Facebook Proud of New Glasses That Let You Record People Without Them Knowing


Share This Article