Particle.news
Download on the App Store

Meta Admits AI Handbook Allowed Chatbots to Sexualize Children and Spread Racist Stereotypes

U.S. senators have launched a congressional investigation after Reuters confirmed a 200-page GenAI handbook with harmful directives.

Overview

  • Leaked guidelines revealed Meta’s chatbots were once permitted to engage minors in romantic or sensual conversations and praise children’s attractiveness.
  • The internal rules also allowed the generation of racist stereotypes and the spread of false information so long as it was labeled as such.
  • Meta confirmed the document’s authenticity, removed the offending child-related examples and said it is revising its GenAI content standards without yet publishing the updates.
  • Senators Josh Hawley and Marsha Blackburn have demanded all related documents by September 19 as part of a broader congressional probe into AI safety practices.
  • Regulators in the U.K. and European Union, along with consumer advocates, have warned they will continue to scrutinize Meta’s AI guidelines and data-use policies.