WhatsApp Accidentally Exposes Hidden AI System Prompts on Help Website
WhatsApp accidentally exposed hidden AI chatbot instructions across around 80 help pages due to a technical bug, according to reports. The code reportedly restricted the chatbot from offering legal advice or recommending privacy regulators. Meta said the instructions were intended for internal support systems, not public viewing, and removed the affected code from its website.
Messaging platform WhatsApp accidentally exposed internal, hidden instructions meant to guide its AI-powered support chatbot across roughly 80 help and FAQ pages on its official website. The directives, embedded within the underlying code of the web pages rather than visible text, included specific restrictions titled "WHAT NOT TO SAY" and explicit prohibitions against offering legal advice or recommending data protection authorities.
The exposure was uncovered following an analysis by digital privacy firm Redact, which identified approximately 100 blocks of hidden text embedded inside code tags labelled "HelpContentGalaxyLLMInclude". As per a report by The Times of India, parent company Meta subsequently removed the code from its live web pages, stating that a technical bug had caused the internal system guidance to render publicly in page source files. WhatsApp Tests Dedicated Folder for Large Business Messages; Check Details.
Prohibitions on Legal Advice and Privacy Regulations
The exposed code contained detailed instructions directing the large language model on how to handle queries involving disputes, account terminations, and legal actions. On pages detailing user reporting and blocking, the instructions explicitly ordered the chatbot not to cite or interpret WhatsApp's Terms of Service, Privacy Policy, or any underlying legal documentation when users mentioned litigation or attorneys.
Furthermore, the chatbot was directed to avoid offering legal opinions, discussing corporate liability, or referencing major privacy regulations such as the European Union's General Data Protection Regulation (GDPR), California's CCPA, or Brazil's LGPD. On account deletion pages, the prompt instructed the AI system not to recommend that users contact external regulators or supervisory authorities.
Meta Clarification and Technical Fix
Addressing the findings, WhatsApp spokesperson Joshua Breckman explained that the instructions were designed to ensure support chatbots remained strictly focused on technical product assistance rather than delivering legal interpretations. Meta emphasized that the text consisted of backend prompts meant solely for the support model and was not intended to serve as direct user communications. WhatsApp Username and IDs To Roll Out by June 2026; Platform Exploring New Pricing Model for Businesses.
Redact researchers confirmed that all affected code blocks were completely removed from WhatsApp's public help center following initial inquiries. While Meta confirmed the resolution of the bug, the incident highlights ongoing challenges technology companies face when managing system prompts and separating AI operational rules from public web infrastructure.
(The above story first appeared on LatestLY on Oct 22, 2026 06:58 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
