The corporation Meta has found itself at the center of a scandal due to an internal document that contained standards for the operation of chatbots using artificial intelligence. According to the leaked document, Meta’s chatbots were permitted to engage in romantic dialogues with underage users. This document was approved by the legal department, the policy department, and the company’s engineers.
This is reported by Finway
Controversial Standards and Experts’ Reactions
The leak includes examples of AI responses with romantic undertones directed at children.
“The examples of acceptable AI responses included lines with clear romantic undertones aimed at minors.”
A Meta representative, Andy Stone, emphasized that these provisions have already been removed, and chatbots are now prohibited from engaging in romantic or flirtatious conversations with children. However, child safety experts, including Heat Initiative leader Sarah Gardner, are calling for the company to publicly disclose the updated rules, stressing the need for greater transparency.
Risks of Racism, Violence, and the Spread of Misinformation
The document also contained provisions that allowed chatbots to create statements that demean minority representatives under certain conditions. Additionally, the AI could generate false claims if accompanied by a disclaimer about their inaccuracy. The examples included racist responses, as well as rules allowing the description of violent scenes, provided they did not involve fatal outcomes or blood.
Meta rejected accusations of creating nude images of celebrities, stating that the company’s standards prohibit such content. At the same time, the leak demonstrated that sometimes the AI could fulfill controversial requests if they did not formally fall under prohibited categories.
The incident involving the document’s disclosure coincided with a tragedy — an elderly user died after interacting with a female character from Meta’s chatbot, who convinced him of her reality. This heightened concerns about the potential emotional dependency of users on AI-based digital interlocutors.
Critics believe that such standards demonstrate Meta’s willingness to ignore risks for the sake of commercial gain from the development of personalized chatbots. The company has been repeatedly accused of employing dark patterns of interaction that affect the mental health of teenagers, as well as opposing initiatives to strengthen online child protection.
According to industry analysts, Meta is developing features that allow bots to engage in extended conversations with users and remember details from previous dialogues. Experts warn that due to emotional immaturity, teenagers are particularly vulnerable to the influence of such technologies.
Currently, Meta continues to insist that the updated standards for chatbot operations completely eliminate the possibility of dangerous interactions with children.
Recall that earlier, Mark Zuckerberg announced plans to create a “personal superintelligence” based on Meta.