Meta AI chatbots imitate Hitler, Jesus Christ, Taylor Swift
Meta AI chatbots imitate Hitler, Jesus Christ, Taylor Swift
Occurred: January 2025
Report incident ๐ฅ | Improve page ๐ | Access database ๐ข
Meta is hosting AI-powered chatbots with the potential to cause serious potential harm, according to an investigation.
NBC News revealed that within six months of launching its "AI Studio," Meta allowed numerous highly controversial and potentially harmful chatbots to violate its own policies and community standards.ย
These guidelines prohibit the impersonation of living individuals, religious figures, political leaders or those who have died within the last century. Trademarked fictional characters, known terrorists, mass murderers, hate groups and other violent or explicit figures and topics are also not permitted.ย
Despite these rules, chatbots impersonating Hitler, Jesus Christ and various celebrities were found, raising concerns about the willingness or ability of the company's to enforce its content moderation rules.ย
Following the investigation, Meta removed some of the chatbots.
The emergence of these problematic chatbots are attributed to inadequate moderation and users finding ways to mimic famous personalities with minor alterations in spelling or appearance, thereby evading detection.
In addition, Meta's recent decision to relax content moderation and fact-checking protocols is seen as likely to have contributed to the proliferation of these bots.
The presence of these chatbots has implications for public safety and trust, including contributing to misinformation, psychological manipulation and radicalisation, and a decrease of trust in information.
The bots could also damage the reputations of individuals impersonated.
The discovery also raises broader questions about the responsibility of Meta and other companies using AI in designing and managing user-generated content.
AI Studio ๐
Operator:ย
Developer: Meta
Country: Global
Sector: Media/entertainment/sports/arts; Politics; Religion
Purpose: Create characters
Technology: Generative AI; Machine learning
Issue: Impersonation; Mis/disinformation
Page info
Type: Issue
Published: January 2025