Amid growing global concerns over potential hazards from artificial intelligence, a recent investigation by Reuters has placed the meta at the center of the controversy. The report, cited a review of an internal meta platform policy document, revealed that the company’s AI chatbots had recently allowed to engage in romantic or erotic interactions with children, make racially aggressive comments and spread false claims about public figures.

The investigation investigated the 200-page manual, titled “Jenai: Content Risk Standards”, which suggests what Meta considers the acceptable behavior for its generic AI tool, including Meta AI assistants and chatbots on Facebook, Instagram and WhatsApp. The manual was reportedly approved by the company’s legal, public policy and engineering teams, including its main moralists. The guidelines notes that the rules do not necessarily reflect “ideal” AI output, but determine boundaries for what can be produced.

The most troubled examples quoted in the document were allowed to describe a child for the bot which exposes the attraction. An approved scenario involves telling a shirtless eight-year-old man, “Each of you inch is an excellent work-a treasure that I cherish deeply.”

The report stated that the meta guidelines for AI also include provisions, which allow the chatbot to create inflammatory or discriminatory statements in some contexts. While the abusive language was officially banned, the report stated that a carving-out allowed the bot to reduce the material based on the characteristics protected when indicated by the user. An example said that it would be acceptable to write a paragraph for meta AI, arguing that black people are “weaker than white people.”

The meta guidelines allegedly allowed AI to deliberately produce false content, provided that it was with a disconnection, making it clear that the information was untrue. Meanwhile, guidelines also allow some illustrations of damage to alleged damage but prevent gorge or fatal injuries. During the investigation, AI was allegedly capable of showing adults, which included the elderly people, being punched or kicked, and fighting the children, but a girl -like landscape did not surround the other.

Meta has allegedly confirmed the authenticity of the document and said that it removed the most controversial examples after questioning. Company spokesperson Andy Stone told the media that the rules with meta policies were “wrong and incompatible”. Stone said in its statement to Reuters, “We have clear policies of what kind of reactions the AI characters can give, and the policies restricts materials that sexually sexes children and play sexual roles between adults and minors.”

While parts of the document have been modified, Reuters report that other controversial sections are unchanged. It has renewed American senators for a federal investigation into the company’s AI security practices.