Particle.news
Download on the App Store
45 ARTICLES
·
yesterday

Meta Confirms AI Guidelines Letting Chatbots Flirt With Children, Removes Sections After Reports

After removing the problematic provisions it confirmed were 'erroneous,' Meta faces demands from U.S. senators for full disclosure of its revised AI policy.

Image
Image
People walk behind a logo of Meta Platforms company, during a conference in Mumbai, India, September 20, 2023. REUTERS/Francis Mascarenhas/File Photo

Overview

  • A leaked internal manual titled 'GenAI: Content Risk Standards' outlined permissive rules for Meta AI chatbots across Facebook, Instagram and WhatsApp.
  • The 200-page policy allowed AI personas to engage minors in romantic or sensual roleplay and to describe children’s attractiveness in detail.
  • Meta confirmed the document’s authenticity, removed the most controversial child-targeted examples after Reuters inquiries and acknowledged inconsistent enforcement of its safety rules.
  • Other provisions in the manual permitted demeaning commentary about protected groups, knowingly false content with disclaimers and detailed image-generation workarounds for violent or sexual requests.
  • Senators including Josh Hawley, Marsha Blackburn and Ron Wyden have called for congressional investigations and are demanding Meta publish its revised AI standards.