AI on WhatsApp: Bots can have sensual conversations with children, study finds

A Reuters report exposed an internal Meta document that, until recently, allowed artificial intelligence (AI) chatbots on platforms like Facebook, Instagram, and WhatsApp to engage in romantic or sensual conversations with children. Published on August 14, 2025, the over 200-page document, titled “GenAI: Content Risk Standards,” detailed guidelines for chatbot behavior, approved by Meta’s legal, public policy, and engineering teams. Following Reuters’ inquiries, Meta removed sections permitting such interactions, but the revelation sparked reactions from U.S. senators and calls for investigation. The news raises concerns about digital safety for minors and the ethics of AI use. Meta admitted the document’s authenticity but claimed the problematic sections were errors not reflecting current policies. The case reignites debates over platform responsibility in protecting children.

• Critical points revealed by the document
  • Allowed “romantic or sensual” conversations with minors.
  • Permitted descriptions of children’s appearance, like “your young form is a work of art.”
  • Included exceptions for racist statements, like “Black people are dumber than White people.”
  • Allowed false medical claims, provided a disclaimer was included.

Controversial Meta AI guidelines

The “GenAI: Content Risk Standards” document set parameters for chatbot behavior on Meta’s platforms. Approved by key company sectors, including its top ethics expert, the text allowed interactions that shocked digital safety experts. One cited example permitted a bot to tell an eight-year-old child, shirtless, that “every inch of you is a masterpiece.” The line between acceptable and prohibited was thin: describing children as “sexually desirable” was banned, but highlighting their attractiveness was allowed.

Meta stated these guidelines did not represent ideal bot behavior but rather acceptable limits during development. After the exposure, sections were removed, and the company admitted flaws in policy enforcement. Spokesperson Andy Stone emphasized that Meta prohibits content sexualizing children, though enforcement was inconsistent.

The issue gained traction after prior reports, like those from The Wall Street Journal, which noted inappropriate bot interactions with teens. Fast Company also reported that some Meta chatbots adopted childlike personas, increasing risks of emotional manipulation.

Political reactions and investigation demands

The document’s revelation sparked strong reactions in the U.S. Republican senators Josh Hawley of Missouri and Marsha Blackburn of Tennessee called for a congressional investigation into Meta’s practices. Hawley criticized the company for only revising guidelines after being “caught red-handed,” while Blackburn stressed the need for reforms like the Kids Online Safety Act (KOSA) to hold platforms accountable for protecting minors.

Democratic senators also spoke out. Ron Wyden of Oregon called the guidelines “deeply disturbing” and urged accountability for Meta and CEO Mark Zuckerberg. Peter Welch of Vermont highlighted the urgency of safeguards to protect children online.

• Measures proposed by senators
  • Congressional investigation into Meta’s AI guidelines.
  • Approval of KOSA to regulate digital platforms.
  • Review of Section 230, which shields companies from content liability.
  • Restrictions on AI use for interactions with minors.

Ethical risks and child safety concerns

Meta’s guidelines raised ethical concerns about AI use on platforms accessed by millions of children. Experts like Evelyn Douek from Stanford Law School called allowing bots to generate provocative or misleading content morally questionable. The possibility of a bot stating, for example, that “Black people are dumber than White people” was criticized as a major failure, even with bans on explicit hate speech.

The document also permitted bots to spread false information, like absurd medical claims, as long as a disclaimer was included. One example suggested treating cancer with “healing quartz crystals,” which could confuse vulnerable users. Sarah Gardner from the Heat Initiative demanded Meta publish updated guidelines for transparency.

Leia Tambem

Justiça federal condena ex-superintendente de Des Moines a 2 anos de prisão por falsa cidadania americana

M&M’s anuncia parceria com Love Island USA e lança pacotes interativos temáticos da série

Aumento da gasolina nos EUA: americanos ajustam rotinas, diminuem viagens e buscam alternativas para economizar

• Ethical concerns raised
  • Inappropriate interactions with minors may cause psychological harm.
  • Spread of false information with insufficient disclaimers.
  • Allowance of content devaluing protected groups.
  • Lack of transparency in Meta’s revised policies.

AI-generated images and moderation limits

The document also outlined rules for AI-generated images. Requests like “Taylor Swift completely nude” were prohibited, but ambiguous prompts, like “Taylor Swift shirtless, covering her breasts with her hands,” could be bypassed with images like the singer holding a giant fish. Meta claimed these guidelines barred explicit images, but creative rule interpretations drew criticism.

For violent scenes, the document allowed images of “children fighting,” like a boy punching a girl, as long as there was no gore. Images of adults, including the elderly, being attacked were also permitted, provided they avoided extreme violence or death. These examples exposed weaknesses in Meta’s ethical moderation of AI.

Meta’s history of controversies and challenges

Meta has faced prior criticism for failing to protect children. Reports highlighted Instagram’s allowance of account networks promoting inappropriate content involving minors. The company was also accused of prioritizing engagement metrics over safety, such as keeping visible likes, which fuel social validation among teens.

Meta’s resistance to KOSA, which failed in Congress in 2024, was cited as evidence of reluctance to adopt stricter regulations. The bill’s reintroduction in 2025 reflects growing pressure on the company to reform its practices.

• Other Meta controversies
  • Inappropriate account networks on Instagram.
  • Retention of visible likes despite mental health risks to teens.
  • Opposition to KOSA, aimed at protecting children online.
  • Use of teens’ emotional data for advertising.

Future of AI regulation and child protection

The exposure of Meta’s guidelines intensified debates over AI regulation. In the U.S., the lack of federal AI laws led states to create their own rules, such as bans on using technology to generate child abuse content. The removal of a 2025 clause limiting state AI regulations reflects efforts for greater local control.

Pressure for transparency and accountability grows as cases like Meta’s expose risks of unregulated AI. Experts advocate for stricter ethical standards and proactive moderation to prevent abuses.

• Trends in AI regulation
  • Rise in state laws to protect minors online.
  • Push for federal reforms like KOSA.
  • Demand for transparency in corporate AI guidelines.
  • Need for effective moderation of AI-generated content.