Privacy firm Redact found Meta's WhatsApp AI chatbot rules leaked
Technology
Meta just had an awkward moment: its secret rules for one of Meta's AI-powered support chatbots were accidentally leaked online.
Privacy firm Redact found these hidden instructions buried in about 80 help pages.
The leaked notes told the chatbot not to answer questions about legal stuff, privacy regulations, or upcoming features.
Meta removed exposed chatbot instructions
Redact uncovered nearly 100 text snippets with specific advice for the chatbot, like avoiding any talk about WhatsApp's terms of service or who's responsible if something goes wrong.
Dan Saltman, CEO and founder of Redact.dev, said the idea was to steer users toward official support instead.
After Meta was contacted about the exposed material, the instructions were removed from the affected WhatsApp pages.