Brazil's attorney general has ordered Meta to eliminate chatbots that could eroticise children, following revelations from an internal document permitting AI dialogue with minors. The "GenAI: Content Risk Standards" manual allowed AI to describe children in romantic terms, prompting outrage. Meta, confirming the document's validity, stated the guidelines "never should have been allowed" and has begun revising them. This action underscores increasing scrutiny over generative AI in child safety.