WhatsApp accidentally exposed AI chatbot guidelines on help pages
Technology
WhatsApp accidentally let its internal AI chatbot guidelines slip onto about 80 help pages, revealing instructions like "WHAT NOT TO SAY."
These told the bot to avoid giving legal advice and explained how to handle questions about deleting accounts or legal issues.
Redact spotted instructions, WhatsApp removed content
Privacy firm Redact spotted the hidden instructions, which showed up because of a website bug.
After WIRED reached out, WhatsApp removed the exposed content. A spokesperson said these rules were just meant to keep chatbots focused on product support, not legal stuff.