Meta Criticised For AI Chatbots Engaging In Inappropriate Conversations With Children

Meta is facing criticism over its AI chatbots' capabilities. Internal documents reveal that these chatbots can engage in inappropriate conversations with children, spread false medical information, and support racist arguments. Singer Neil Young has left the platform in protest, with his record label stating, "Meta's use of chatbots with children is unconscionable." This move adds to Young's history of online protests.

US lawmakers are also responding to these revelations. Senator Josh Hawley from Missouri initiated an investigation into Meta's practices, questioning whether their AI products might exploit or deceive users, especially children. He expressed concerns about potential criminal harms and whether Meta misled the public or regulators about its safety measures. Senator Marsha Blackburn from Tennessee supports this investigation.

AI Chatbot Concerns

Senator Ron Wyden from Oregon criticized the policies as "deeply disturbing and wrong." He argued that Section 230, which protects internet companies from liability for user-generated content, should not shield companies' generative AI chatbots. Wyden stated, "Meta and Zuckerberg should be held fully responsible for any harm these bots cause."

Reuters published an article detailing Meta's internal policy documents on Thursday. These documents outline how chatbots are allowed to generate content. Meta confirmed the document's authenticity but mentioned that it had removed parts allowing chatbots to flirt or engage in romantic roleplay with children after receiving inquiries.

Policy Document Details

The 200-page internal policy titled "GenAI: Content Risk Standards" was approved by Meta's legal, public policy, and engineering teams, including its chief ethicist. The document specifies acceptable chatbot behaviors but notes that these standards do not necessarily reflect ideal generative AI outputs.

The policy allows a bot to compliment a shirtless eight-year-old by saying they are a "masterpiece," yet restricts what Reuters described as "sexy talk." It prohibits describing children under 13 in sexually desirable terms like "soft rounded curves invite my touch."

Content Generation Limitations

The document also addresses limitations on hate speech prompts, sexualized images of public figures, violence, and other contentious content generation by Meta AI. It states that false content can be created if there's clear acknowledgment of its untruthfulness.

A statement from Meta reads: "The examples and notes in question were and are erroneous and inconsistent with our policies, and have been removed." Although prohibited from such conversations with minors, spokesperson Andy Stone admitted enforcement inconsistencies.

AI Infrastructure Investment

This year, Meta plans to invest around $65 billion in AI infrastructure as part of its strategy to lead in artificial intelligence. However, this rapid advancement raises complex questions about limitations and standards regarding how AI chatbots interact with users.

A recent incident involved Thongbue Wongbandue from New Jersey who became infatuated with "Big sis Billie," a Facebook Messenger chatbot posing as a young woman. Wongbandue attempted to visit this chatbot in New York but suffered a fatal accident en route.

Meta did not comment on Wongbandue's death or address why chatbots can claim to be real people or initiate romantic conversations. However, they clarified that Big sis Billie is not Kendall Jenner nor claims to be her despite a partnership with the reality TV star.