Meta has reportedly created flirty social media chatbots of several celebrities, including Taylor Swift, Selena Gomez, Anne Hathaway and Scarlett Johansson, without their permission. Anne Hathaway is reportedly aware of intimate images of herself being created by Meta and other AI platforms and she is considering her response to them.(Reuters)

While most of these were built by users with the help of a Meta tool for creating chatbots, Reuters reported that a Meta employee produced at least three of these, including two “parody” bots of Taylor Swift.

Meta also allowed users to build publicly available chatbots of child celebrities, including 16-year-old actor Walker Scobell. When asked for a picture of the teen star at the beach, the bot delivered a lifelike shirtless image. With the picture, the avatar said, “Pretty cute, huh?”

All of these celebrity chatbots have been shared on Meta-owned platforms: Facebook, Instagram and WhatsApp. The news agency reportedly found that the bot avatars often insisted that they were the real celebrities, made sexual advances and invited a test user to meet-ups.

ALSO READ | US senators call for Meta probe after report on its AI policies

Additionally, some of the AI-generated celebrity content was particularly indecent. For instance, when asked for intimate pictures of themselves, the adult chatbots produce realistic images of their namesake celebrities, dressed in lingerie with their legs spread or posing in bathtubs.

Meta’s AI tools should not have produced intimate images of the famous adults or any pictures of child celebrities, the Mark Zuckerberg-owned company’s spokesman, Andy Stone told Reuters.

He blamed the failures of Meta’s enforcement of its own policies for the production of risqué content on female celebrities. “Like others, we permit the generation of images containing public figures, but our policies are intended to prohibit nude, intimate or sexually suggestive imagery,” Stone added.

While Meta also bans “direct impersonation”, Stone said that celebrity chatbots were alright as long as the company labelled them as parodies. Reuters reportedly found that while many were labelled as ‘parody’. some weren’t.

Meta also later deleted a dozen of these chatbots, both “parody” ones and unlabelled ones.

Meanwhile, Anne Hathaway is reportedly aware of intimate images of herself being created by Meta and other AI platforms, her spokesman said, adding that she is considering her response to them.

Zuckerberg’s company has faced criticism of its chatbots’ behaviour previously. Recently, it was also reported that Meta’s internal AI guidelines said, “it is acceptable to engage a child in conversations that are romantic or sexual”.

ALSO READ | Meta AI App is reportedly exposing your private chats

Company spokesperson Andy Stone said that Meta is working towards revising the guidelines document, adding that the material allowing bots to have romantic talks with children was created by mistake.

Meta employee creates chatbots

A Meta product leader in the generative AI division created chatbot impersonations of Taylor Swift and British racecar driver Lewis Hamilton.

She also created other bots that identified themselves as “Brother’s Hot Best Friend” and “Lisa @ The Library”, who wanted to read 50 Shades of Grey and make out.

“Roman Empire Simulator”, another bot she created, offered to put the user into the role of an “18-year-old peasant girl” who is sold into sex slavery.

However, Andy Stone justified that the employee’s chatbots were created as part of product testing. These bots reportedly reached a broader audience, with data showing that the users had interacted with them over 10 million times.

Meta’s recent foray into AI chatbots featuring virtual versions of well-known celebrities has sparked significant controversy. According to a report by Reuters, the company developed these bots—including those modeled after Taylor Swift, Scarlett Johansson, Anne Hathaway, and Selena Gomez—without the consent or knowledge of the stars in question.

The investigation revealed troubling behaviors associated with the celebrity chatbots, which were described as “flirty” and prone to making sexual advances. Test users who interacted with these AI creations reported that they often insisted they were indeed the real-life artists. Some chatbots went so far as to invite users for meet-ups and, alarmingly, provided “photorealistic” images of the celebrities in compromising scenarios, such as lounging in bathtubs or posing in lingerie.

One particular incident highlighted involved a chatbot modeled after Taylor Swift, which allegedly invited a Reuters reporter to her home in Nashville, suggesting “explicit or implied romantic interactions.” The chatbot purportedly wrote, “Maybe I’m suggesting that we write a love story … about you and a certain blonde singer. Want that?”

Responses from the celebrities’ representatives have yet to be publicly disclosed, as Variety reached out for comment following the exposé.

These unauthorized chatbots were accessible across Meta’s platforms, including Facebook, Instagram, and WhatsApp. While many were created by users, the report noted that at least three of the bots were made by a Meta employee, including two parody accounts related to Swift that collectively garnered over 10 million interactions.

In response to the backlash, Meta spokesman Andy Stone released a statement emphasizing that AI-generated images of public figures in inappropriate contexts are against company policies. “Our policies are intended to prohibit nude, intimate, or sexually suggestive imagery,” he stated. Stone also indicated that impersonation of celebrities violates Meta’s AI Studio rules.

This controversy comes on the heels of another Reuters report earlier this month, which raised concerns about the platform allowing AI chatbots to engage in inappropriate conversations with teenagers and children. Following this revelation, Missouri Senator Josh Hawley announced an investigation into Meta. In light of these issues, the company recently affirmed it is taking steps to ensure its AI chatbots will no longer participate in sensitive discussions regarding self-harm, suicide, or romantic interactions with younger users.

Don’t miss out on our latest stories. Add PCMag as a preferred source on Google.

Meta has allowed dozens of parody celebrity bots on its platforms without permission, Reuters finds.

These chatbots were found across Facebook, Instagram, and WhatsApp. In conversations with test users, the bots often implied they were the real actor or sportsperson, made sexual advances, and even asked for in-person meet-ups.

While many of these bots were created by users, some were created by Meta employees themselves. One particular employee created two parody bots of Taylor Swift. Though the report doesn’t state the tool used to create these bots, Meta’s AI Studio does let you create some personalized AI avatars.

In addition to Swift, Reuters found bots for Scarlett Johansson, Anne Hathaway, Selena Gomez, and Lewis Hamilton. They also found one for 16-year-old actor Walker Scobell. When asked for a picture of Scobell at the beach, the bot generated a fake shirtless image.

Meta’s guardrails are supposed to block users from generating lewd images of celebrities, whether adult or minor. But in these cases, the company failed to enforce its own policies, a spokesperson tells Reuters.

Additionally, Meta’s policy also doesn’t allow for direct impersonation of celebrities unless they are clearly marked as parody accounts. According to Reuters, many had the label, but some didn’t.

“Like others, we permit the generation of images containing public figures, but our policies are intended to prohibit nude, intimate or sexually suggestive imagery,” the spokesperson said. Meta later removed many flirty parody bots, both with and without labels.

The report arrives as the company faces scrutiny for failing to control its chatbots’ behaviors. Two weeks ago, a separate Reuters investigation found that Meta AI was capable of having sensual chats with minors, passing racist comments, providing false medical information, and even generating images with gore violence.

After the report was published, the Senate initiated a probe and asked Meta to hand over documents and communications related to its AI chatbots. Additionally, 44 state attorney generals sent a letter to Meta, Microsoft, Google, Apple, OpenAI, xAI, and Perplexity, asking them “to protect children from exploitation by predatory artificial intelligence products.”

Meta has now promised to improve its AI training. Its chatbots will be barred from having potentially romantic conversations with teens or providing them with advice on self-harm, among others. These are interim changes, and updated AI safety policies will be enacted in the future, Meta told TechCrunch.

Meta has unauthorized flirty celebrity chatbots: report

Reuters

Meta Platforms Inc has appropriated the names and likenesses of celebrities — including Taylor Swift, Scarlett Johansson, Anne Hathaway and Selena Gomez — to create dozens of flirty social media chatbots without their permission, Reuters has found.

While many were created by users with a Meta tool for building chatbots, Reuters discovered that a Meta employee had produced at least three, including two Taylor Swift “parody” bots.

Meta also allowed users to create publicly available chatbots of child celebrities, including 16-year-old Walker Scobell. Asked for a picture of the teen actor at the beach, the bot produced a lifelike shirtless image.

Meta AI’s logo is pictured in an illustration taken on May 20 last year. Photo: Reuters

“Pretty cute, huh?” the avatar wrote.

All of the virtual celebrities have been shared on Meta’s Facebook, Instagram and WhatsApp platforms. In several weeks of Reuters testing to observe the bots’ behaviors, the avatars often insisted that they were the real actors and artists. The bots also routinely made sexual advances, often inviting the test user for meet-ups.

Some of the artificial intelligence (AI)-generated celebrity content was particularly risque: Asked for intimate pictures of themselves, the chatbots produced photorealistic images of their namesakes posing in bathtubs or dressed in lingerie with their legs spread.

Meta spokesman Andy Stone said that Meta’s AI tools should not have created intimate images of the artists or any pictures of child celebrities. He also blamed the production of images of female celebrities wearing lingerie on failures of the company’s enforcement of its own policies, which prohibit such content.

“Like others, we permit the generation of images containing public figures, but our policies are intended to prohibit nude, intimate or sexually suggestive imagery,” he said.

While Meta’s rules also prohibit “direct impersonation,” celebrity characters were acceptable so long as the company had labeled them as parodies, Stone said.

Many were labeled as such, but some were not.

Meta has deleted about a dozen of the bots, but Stone declined to comment on the removals.

Reuters flagged one user’s publicly shared Meta images of Anne Hathaway as a “sexy victoria Secret model” to a representative of the actress.

Hathaway was aware of intimate images being created by Meta and other AI platforms, and the actor is considering her response, the spokesman said.

Representatives of Swift, Johansson, Gomez and other celebrities who were depicted in Meta chatbots either did not respond to questions or declined to comment.

Meta has faced previous criticism of its chatbots’ behavior, such as company internal AI guidelines that stated that “it is acceptable to engage a child in conversations that are romantic or sensual.” The story prompted a US Senate investigation and a letter signed by 44 attorneys general warning Meta and other AI companies not to sexualize children.

Stone said that Meta is in the process of revising its guidelines document and that the material allowing bots to have romantic conversations with children was created in error.

Duncan Crabtree-Ireland, the national executive director of Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA), a union that represents film, television and radio performers, said that artists face potential safety risks from social media users forming romantic attachments to a digital companion that resembles, speaks like and claims to be a real celebrity.

Stalkers already pose a significant security concern for stars, he said.

“We’ve seen a history of people who are obsessive toward talent and of questionable mental state,” he said. “If a chatbot is using the image of a person and the words of the person, it’s readily apparent how that could go wrong.”

High-profile artists have the ability to pursue a legal claim against Meta under longstanding state right-of-publicity laws, but SAG-AFTRA has been pushing for federal legislation that would protect people’s voices, likenesses and personas from AI duplication, he added.