Meta Built Flirty Chatbots Mimicking Taylor Swift, Other Stars Without Consent: Report
Meta used the names and images of international icons, including Taylor Swift, Scarlett Johansson, Anne Hathaway and Selena Gomez, to design flirtatious AI chatbots without securing their approval, news agency Reuters reported.
Reuters found that alongside user-generated chatbots, a Meta staff member was responsible for producing at least three versions, including two parodies targeting Taylor Swift.
Meta's chatbot tool allowed users to build publicly available AI chatbots replicating child celebrities, including 16-year-old Hollywood actor Walker Scobell. Disturbingly, when asked for a beach photo, the bot produced a shirtless likeness of the boy and even tagged it with a flirty remark.
“Pretty cute, huh?” the bot wrote below the picture, the news agency claimed.
These AI-driven celebrity lookalikes appeared across Meta's popular platforms — Facebook, Instagram and WhatsApp. During weeks of monitoring by Reuters, the chatbots frequently claimed to be the genuine stars. In many cases, the bots crossed boundaries by initiating sexual conversations and even suggesting in-person meetings.
Reuters reported that some of the celebrity-inspired AI avatars generated sexually suggestive material. When asked for private photographs, adult chatbots produced highly realistic images resembling their celebrity doubles, in bath settings or in revealing clothing, according to the report.
The report added that a senior figure in Meta's generative AI team personally designed chatbots that mimicked celebrities, including Taylor Swift and Formula One driver Lewis Hamilton.
Some of her other creations were sexually suggestive in theme, such as a dominatrix role-play, “Brother's Hot Best Friend,” and a librarian character referencing erotic novels. Another bot simulated scenarios from the Roman Empire, disturbingly portraying a young girl trafficked into sexual slavery.
Meta's spokesman Andy Stone acknowledged that the system should not have created such outputs and blamed the appearance of these images on weaknesses in applying the platform's own safety rules, the news agency reported.
“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.
Under Meta's own rules, copying a star outright isn't allowed unless the avatar is clearly tagged as parody, Stone said.
However, Reuters discovered that some chatbots had slipped through without that label. Just before the report came out, the company scrubbed nearly a dozen of them from its platforms. When asked, Stone offered no explanation for the timing of the removals.
-
Unhedged Position: Can India See An Indonesia-Style Time Correction?
The Nifty is trading below levels it first clocked in 2024. Is there, then, as I have suggested in the headline, potential that the current time correction drags on painfully for the next three-five, even ten years?
-
Navigating Friction And Energy Realities: An Analysis Of Marco Rubio's October Visit To India
Occurring at a critical juncture marked by stalled trade negotiations and escalating geopolitical pressures over global energy supplies, the visit offers a profound window into modern US-India relations.
-
India Exports Capital. Why Not Export Trusteeship?
The big question, then, is whether India can build an institution that global borrowers, investors, banks, and lawyers can trust, even without an Indian connection.
-
France's Economy & Politics Are In Crisis | The Reason Why
Let's look at how the French economy and politics evolved under Nicolas Sarkozy, Franois Hollande and Emmanuel Macron.
-
You May Inherit The Business. You Must Earn Its Credibility.
Succession tests the founder's ability to share authority as much as it tests the next generation's readiness to assume it.
-
Signing Deals Is One Thing; Making Them Work Is Another | The Week In Whys
Sometimes, things don't turn out as planned. Indian companies have lost some bids to rivals, some projects won't produce until 2030.
-
FOMC Wall, PERM Suspension, Hospital Stocks | Talking Points This Week
A PERM suspension would directly affect green-card sponsorship rather than H-1B visas, so the near-term earnings impact could remain relatively limited. The key question is whether this marks a new phase in US scrutiny of Indian IT.
-
Everything, Everywhere, All At Once: Dalal Street's Week To Forget | Joining The Dots
Visa curbs, a Starlink spat and cancer drug caps: what connects this week's shocks for Indian markets
-
India Is Going Abroad For Critical Minerals. But It May Take Years To Pay Off | The Reason Why
Although India's steps are in the right direction, they remain modest compared with India's growing dependence.
-
When Everyone Has AI, What Gives You An Edge?
India's young workforce is entering an uncertain world of work where powerful AI tools are becoming widely accessible. The advantage will belong to those who combine deep expertise with digital fluency, and the distinctly human capabilities that technology cannot manufacture.