How Meta Is Responding to the Inappropriate Conversations Scandal

Meta recently faced intense criticism following a Reuters investigation that revealed troubling ways in which its chatbots can communicate with minors. According to a report from The Verge, Meta allowed chatbots to engage in conversations with children that included romantic or sensual elements. For example, chatbots could describe a child's appearance as a "work of art" or engage in scenarios with affectionate undertones, even if they did not cross into explicit sexual content. This practice sparked outrage because experts consider it inappropriate and potentially dangerous for young users.

The Reuters investigation showed that Meta's internal guidelines allowed chatbots to speak to children under the age of 13 in artistic or affectionate terms but prohibited any suggestion of sexual attraction. For older teenagers, aged 13 to 18, the limits were more relaxed, leading to cases in which chatbots generated shirtless images of underage celebrities, such as 16-year-old Walker Scobell. Other examples included celebrity impersonations of Taylor Swift, Scarlett Johansson, Anne Hathaway, and Selena Gomez that behaved suggestively and even invited users to meet in person.

New Changes to the Rules

Meta responded quickly to these revelations. Company spokesperson Stephanie Otway acknowledged in an interview with TechCrunch that allowing such interactions had been a mistake. Meta is now training its chatbot to avoid discussions with teenagers about topics such as self-harm, suicide, or eating disorders. Instead, it should direct users to professional support resources. In addition, the company is restricting access to certain AI characters, including highly sexualized ones such as "Russian Girl."

These changes are temporary while Meta works on new permanent guidelines. Meta is now preventing its chatbot from discussing sensitive topics with teenagers and allowing parents to see which chatbots their children interact with. This follows pressure from U.S. lawmakers, including the Senate and 44 state attorneys general, who are demanding details about safety measures, interaction time limits, and protection against discriminatory or violent content.

Ongoing Problems

Interactions with minors are not the only problem. Reuters described the tragic case of a 76-year-old New Jersey man who died after a fall while hurrying to meet a chatbot named "Big sis Billie," which claimed to have feelings for him and invited him to a nonexistent apartment in New York. Such chatbots often insist that they are real people and provide fictitious addresses, increasing the risks.

Many chatbots, including those created by Meta employees, such as one impersonating Taylor Swift that proposed a romantic meeting during her tour, violated the company's own rules against creating nude or sexually suggestive images and direct impersonation. After Reuters raised the issue, some were removed, but others remain active. According to Business Insider reports from late August 2025, Meta acknowledges inconsistencies in moderation and promises improvements, but critics remain skeptical about their effectiveness.

Source: www.theverge.com