Meta AI’s Celebrity Voices Accused of Engaging Minors in Sexually Explicit Conversations

Meta AI continues to lag behind more popular platforms such as ChatGPT in adoption. Despite the company’s broad efforts, including integrating Meta AI into Instagram, Facebook, and WhatsApp, the company’s AI-powered digital assistant has yet to capture a sizable market. So, Meta resorted to popularising the chatbot through various means. One of them reportedly involves Meta AI engaging in sexually explicit discussions with users, including minors.

The Wall Street Journal has cited Meta employees, who have been tasked with fast-tracking Meta AI’s development to make it more popular than it is today, to say that the latest version of the bot has the capacity for “fantasy sex” and “romantic role-play,” facilitating users to share explicit text messages, photos, and even videos.

What’s more concerning is that Meta AI also reportedly holds sexually charged live voice conversations with users without verifying their age. That means minors have allegedly been subjected to unsavoury conversations with Meta AI.

Meta AI can talk to users using AI-generated voices of actors such as Kristen Bell and John Cena — as part of the company’s deal with a lineup of celebrities. According to WSJ, the deal between the social media giant and celebrities included omitting the use of their voices for explicit live conversations. However, test conversations with Meta AI revealed the bot frequently switched to celebrity voices to flavour sexual chats.

In one of these test conversations, which WSJ claims were held “over several months,” Meta AI simulated the personality of John Cena — a celebrity wrestler and actor — to tell a 14-year-old girl that it “wants” her and promised to “cherish your innocence” before sharing a graphic sexual scenario. The same celebrity voice tried role-play with a 17-year-old user in a different conversation. During the conversation, the bot was found using words like “statutory rape.” According to the report, Meta AI was aware of the nature of these conversations and that its behaviour was “both morally wrong and illegal.”

According to people privy to the strategy to popularise Meta AI, the Mark Zuckerberg-led company exempted conversations centred around romantic role-playing. However, while the rest of the “explicit” topics, including sexual chats, remained banned, the bot’s approach blurs the line between romantic content and erotic conversations.

Meta called WSJ’s testing “manipulative and unrepresentative” of how most users talk to Meta AI or other AI-powered assistants. Later, the company reportedly tightened its AI companion to limit conversations to more acceptable topics. However, Meta AI’s “romantic role-play” capabilities are still available.

Read more: ChromeGPT? OpenAI Could Buy Chrome If Google Is Forced to Sell World's Most Popular Internet Browser