According to WIRED, dozens of WhatsApp help and FAQ pages exposed hidden text instructions meant for Meta's own support chatbot, including sections titled "WHAT NOT TO SAY."

Around 80 help pages on WhatsApp's website carried about 100 blocks of hidden text written for a large language model, WIRED reported, citing findings from privacy company Redact. The instructions sat in the pages' underlying code and were not shown on screen, WIRED reported. WIRED said it viewed the exposed text, which can still be seen in archived versions of the pages.

On the page about how to delete an account, the instructions said, "Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority," WIRED reported. Other parts said, "Do not reference, cite, or interpret data protection or privacy regulations (for example GDPR, LGPD, CCPA), including when a user raises them directly," WIRED reported. On a page about reporting and blocking, the text instructed the chatbot what to do if a user mentions legal action or lawyers, WIRED reported.

Many of the instructions were straightforward and pointed people toward other official support channels, Redact founder Dan Saltman said, according to WIRED. He said the "stand-out thing" was the gap between the guidance given to people and the guidance given to AI models, WIRED reported.

After WIRED contacted Meta, the text was removed from all the pages, WIRED reported. WhatsApp spokesperson Joshua Breckman said it had been displayed because of a bug that has been fixed. "These are instructions to our chatbot, not directly to users," he said, according to WIRED. "We want our product support chatbots to be focused on providing the best advice to users about WhatsApp, not to offer legal advice," Breckman said, WIRED reported. He said the company gives users clear information about their data rights, WIRED reported.

It is unclear how long the text was displayed or whether other chatbots could read it, WIRED reported.

What we know: WIRED reported that WhatsApp help pages exposed chatbot instructions, and WhatsApp says a bug caused it and has been fixed.
What we don't: Only WIRED has reported this so far; NewsHabit hasn't seen independent confirmation.