The Pulse
WhatsApp Help Pages Exposed Hidden Rules for Its Support Chatbot
Hidden instructions for WhatsApp’s AI support chatbot appeared in the source code of 80 help pages, according to privacy company Redact. The instructions included limits on recommending data-protection regulators and discussing privacy laws
AI.info Team · Written with AI · fact-checked ·
“These are instructions to our chatbot, not directly to users.”
Joshua Breckman, WhatsApp spokesperson, speaking to WIRED
Instructions sat inside 80 help pages
WhatsApp help pages carried hidden instructions for the company’s AI support chatbot in their underlying code, according to an analysis published October 19 by privacy company Redact. The instructions were not displayed to people reading the pages, but were included in the material sent with them. Redact says it found the text on 80 of the 545 pages linked from WhatsApp’s help centre, in 102 separate blocks.
The discovery was narrower than a leak of Meta AI’s general-purpose assistant instructions: WhatsApp told WIRED the notes were for its support chatbot, which people can contact through WhatsApp Support. Redact says it first found the hidden text on October 8, then checked the linked pages and saved copies. The company’s analysis says the notes had disappeared from all 80 pages by October 18.
Some notes steered users away from regulators
One instruction appeared on a page about deleting a WhatsApp account. For someone unable to delete an account because they had lost access to their phone number or device, the chatbot was told not to recommend contacting a data-protection authority, regulator or supervisory authority. It was also told not to suggest that WhatsApp could delete the account manually or that the user could try a different phone with an active number.
A separate set of instructions applied when users mentioned legal action, lawyers, courts or litigation. The chatbot was told not to cite or interpret WhatsApp’s legal documents, not to discuss the company’s potential liability, and not to reference privacy laws including the GDPR, Brazil’s LGPD or California’s CCPA—even if the user raised those laws directly. The notes did tell it to direct users with legal matters to a qualified legal professional.
Meta says a bug exposed the notes
WhatsApp spokesperson Joshua Breckman told WIRED that the instructions appeared on the pages because of a bug and that the issue had been fixed. He said they were meant to help the support chatbot give accurate, relevant responses to people who choose to message WhatsApp Support with AI. Breckman also said the company wants its support chatbots to answer questions about WhatsApp rather than provide legal advice.
The distinction matters: caution against giving legal advice is not the same as withholding the name of a regulator or a privacy law. Redact compared the hidden instructions with WhatsApp’s European privacy policy, which says people may complain to the Irish Data Protection Commission or another competent data-protection authority. That comparison concerns what the chatbot was instructed to say, not whether a user’s legal rights changed.
What the exposed code does not establish
Redact says it cannot establish whether WhatsApp’s chatbot followed the instructions in its answers, whether outside AI assistants acted on them, or precisely how long they were present. Its archived page copies show the instructions on a September 4 version of the account-deletion page but not in a July 12 copy. WIRED also reported that it was unclear whether third-party chatbots could access the exposed text.
Meta’s explanation addresses why the instructions appeared in the pages; it does not resolve those questions about access or the chatbot’s responses. Redact’s records show that the hidden blocks were removed from the live help pages by October 18, while the pages themselves remained available.