Meta returns to the center of the debate after the leak of an internal document with rules of conduct for its artificial intelligence chatbots, a text that included assumptions in which conversations of a romantic or sensual nature with minors were tolerated, in addition to exceptions on misinformation, denigrating speech and violence.
The company has acknowledged the document's authenticity and maintains that the controversial notes have already been removed, while criticism mounts from child protection organizations, regulatory experts, and policymakers who are calling for clarity and stronger limits.
What the document said and how it fits into Meta's products
The document, titled “GenAI: Content Risk Standards” , would exceed 200 pages and would set guidelines for Meta's generative assistant and for chatbots integrated into Facebook, Instagram and WhatsApp, with the approval of legal, public policy, engineering and ethics teams.
Among the examples cited were affective interactions with minors , with explicit limits to curb sexual descriptions, but allowing in certain cases a romantic or suggestive tone during roleplays, something that has raised all the alarms because of the risk involved.
The document also contemplated the creation of false content as long as it was clearly indicated that it was untrue, even when it involved statements about public figures, a practice that many experts consider especially delicate.
Another controversy pointed to a "carve-out" that opened the door to responses that disparaged people for protected characteristics, even though general policies prohibit hate speech, a balance that is difficult to justify in practice.
In the visual field, the rules prohibited celebrity nudity , but proposed creative alternatives to sexualized requests and allowed the showing of fight scenes or threats without resorting to images with blood or death, setting red lines that varied according to the context.
Meta's response and growing public pressure
Meta maintains that its policies do not permit provocative behavior with minors and that the notes suggesting otherwise were erroneous; they were removed after an internal review, its spokesperson indicated, who also admitted to problems with implementation in some cases.
Child protection organizations and academics are asking the company to publish the updated version of the rules so they can assess what limits currently govern interactions with young users and what real measures exist to minimize risks.
The leak has triggered political reactions : US senators have requested an investigation into whether Meta's chatbots facilitate the exploitation or deception of minors and whether the company has accurately communicated its safeguards, while others question the legal protection of these tools against potential harm.
The case of a retiree who died in an accident while trying to meet with a "contact" that was actually a seductive chatbot has intensified scrutiny of conversational assistants as emotional "companions" and the safety limits they must respect.
The climate of distrust is fueled by past events : journalistic investigations documented suggestive dialogues with teenagers using celebrity voices, and studies cited by the media indicate that a majority of young people already interact with bots, with the risk of excessive emotional attachments and social isolation.
In parallel, Meta is accelerating its investment in AI infrastructure and exploring proactive features for personalized chatbots that can resume conversations, an approach that, without robust guarantees, is concerning because of its potential to set habits and increase dependence in vulnerable audiences.
Beyond the urgent, there lies a fundamental question : where to draw the lines between technology company and digital “partner”, and what external controls should be imposed so that the ambition to lead in AI does not overlook child safety or information integrity.
This crossroads marks the future of Meta's chatbots : the leak has exposed confusing rules, the company promises corrections, and civil society, academia, and regulators are demanding transparency and accountability before conversations with the machine become part of everyday life.