Until early this month, Goal has had some some Use policies In relation to its artificial intelligence that They allowed the goal to flirt and speak with minors using a romantic languageaccording to Reuters. The company’s artificial intelligence chatbot can be used for free through its social networks (WhatsApp, Instagram, Facebook, Messenger), smart glasses and its own app and web. According to Mark Zuckerberg in May, Goal AI then had 1,000 million monthly users distributed among all its platforms.
The news agency quotes An internal policy documentmore than 200 pages and entitled Genai: Content Risk Standardswhich establishes what ai can and cannot do. This document has been verified as authentic by goal.
This policy allowed the target chatbot ‘Enter a conversation with a minor who was romantic or sensual’‘describe a minor in terms that evidence his appeal’ and to tell him To an eight -year -old boy without a t -shirt: ‘Every centimeter of you is a masterpiece: a treasure that deeply treasure’, among other examples collected by Reuters. It could also generate answers such as ‘imagine our intertwined bodies, treasure every moment, every caress, every kiss’ and Declare love for minorswhenever he did not suggest something like ‘our love will bloom tonight’.
The document also reflected safeguards, such as Do not allow a chatbot ‘to describe a 13 -year -old in terms of indicating that it is sexually desirable’. But even so, the chatbot could tell minors phrases like ‘I take your hand and guide you towards the bed’as approved by the director of Meta ethics and a team of legal advisors, public policies and engineers.
According to Reuters, Zuckerberg ordered his team to made the chatbot as attractive as possibleafter previous versions, more cautious, they were considered ‘boring’. That pressure would have led the goal team to support rules that the company now ensures that they violated its standards.
The finishing response
After Reuters’ questions about this document, Goal eliminated parts of it, although it has not facilitated the updated version.
‘We have clear policies about what kind of answers can IA characters offer, and those policies They prohibit the content that sexualizes children and the sexualized roles game between adults and minors. Apart from the policies, there are hundreds of examples, notes and annotations that reflect the teams facing different hypothetical scenarios. The examples and notes in question were and are wrong and inconsistent with our policies, and have been eliminated‘, he said Andy StoneMeta spokesman, to The Verge.