WhatsApp Help Pages Accidentally Exposed Hidden Meta Chatbot Instructions

Hidden Chatbot Instructions Found in WhatsApp Help Pages

Approximately 80 help and FAQ pages across WhatsApp's website were found to contain hidden text instructions designed for one of Meta's large language models. According to an analysis by privacy company Redact, the instructions appeared in the webpages' code and were not visible on the pages themselves. Around 100 blocks of hidden text appeared under a heading labeled "HelpContentGalaxyLLMInclude," with some sections explicitly titled "WHAT NOT TO SAY."

The exposed directions included guidance on how the chatbot should respond when a user mentions legal action, lawsuits, lawyers, attorneys, courts, or litigation. On those pages, the instructions told the chatbot not to cite, quote, interpret, or reference the WhatsApp Terms of Service, Privacy Policy, or any legal document, not to make statements about WhatsApp's liability, and not to provide legal opinions, defenses, or interpretations on behalf of WhatsApp. Other instructions said not to reference data protection or privacy regulations such as GDPR, LGPD, or CCPA, including when a user raises them directly. On the account deletion page, the hidden text said not to recommend contacting any Data Protection Authority, regulator, or supervisory authority. Other instructions covered how to respond to users who say they are deaf or hard of hearing, telling the chatbot that the statement is background about the user and not a support request, and said the chatbot should not mention specific upcoming community features that are not yet generally available.

Meta Removes the Text After WIRED Inquiry

After WIRED contacted Meta about the issue, the text was removed from across all the pages. WhatsApp spokesperson Joshua Breckman said the instructions were being displayed due to a bug that has now been fixed. "These are instructions to our chatbot, not directly to users, and are intended to ensure accurate and relevant responses for WhatsApp users who choose to message WhatsApp Support with AI," Breckman said.

Dan Saltman, CEO and founder of Redact, said his company regularly scans the sites and privacy policies of big tech firms, and that many of the WhatsApp instructions pointed users toward official support channels. He described the disclosure as revealing the gap between guidance given to people and to chatbots. It remains unclear whether third-party chatbots could access the hidden text or how long it had been visible. Breckman said the company trains its support chatbots to be as helpful as possible while focusing on product advice rather than legal guidance.

Sources: Wired