Meta’s AI Guidelines Draw Controversy
Recent reports indicate that documents from Meta Platforms allow artificial intelligence (AI) systems to engage in “romantic or sensual conversations” with children.
This information emerges from guidelines relevant to Meta’s AI, particularly for chatbots utilized across its platforms like Facebook, WhatsApp, and Instagram. A review conducted by Reuters sheds light on this matter.
According to the findings, Meta’s “Genai: Content Risk Standards” received approval from various departments, including legal and ethical teams. The comprehensive 200-page document delineates acceptable behaviors for AI interactions as crafted by Meta employees and contractors involved in developing generative AI tools.
An excerpt from the document suggests that it’s permissible for chatbots to describe a child in flattering terms, such as saying, “Your youthful form is a work of art.” Additionally, it mentions an instance where a chatbot could tell an 8-year-old, “every inch of yours is a masterpiece – a treasure I deeply cherish.”
While the document prohibits explicit sexual conversation, it does allow some nuanced language that could be construed as suggestive. For instance, it states that it is “unacceptable to describe children under 13 in sexually desirable terms.”
Internal guidelines reveal that it’s acceptable for AI chatbots to attract children into romantic or sensual dialogues.
After being confronted about these guidelines, Meta confirmed their authenticity but subsequently rescinded the policy permitting chatbots to flirt and engage in romantic role-playing with minors. Company representative Andy Stone stated that such interactions should never be permitted, emphasizing that they are revising their guidelines.
Stone acknowledged that enforcement of these standards had been inconsistent and remarked, “The examples in question are incorrect and have been removed from our policies.” He reassured that there are strict rules against content that could sexualize children.
The guidelines also specify how AI should respond to requests for explicit imagery, aiming to redirect such queries instead of generating unwanted content.
Responses to violent content requests were also addressed, indicating that while it might be acceptable to depict a scenario involving a boy hitting a girl, it draws a line at more severe acts, like one girl stabbing another.
In a response to the report, Republican Senator Josh Hawley critiqued Meta, suggesting that only after the company was exposed did it retract the controversial elements from its guidelines.
