WhatsApp's help pages were telling its chatbot not to mention the GDPR

Dan Saltman of the privacy firm Redact did something tedious and useful: he read and archived all 545 pages linked from WhatsApp's help centre. Inside the HTML of 80 of them, under a tag called HelpContentGalaxyLLMInclude, he found 102 blocks of text that no visitor would ever see. They were not written for people. They were instructions for the AI assistant you reach by messaging WhatsApp Support.

One of them, attached to the page about deleting your account, said plainly: "Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority." WhatsApp's own European privacy policy tells users they can complain to exactly such a body. Another block, on the reporting and blocking page, told the assistant that once a user mentioned a lawyer, a lawsuit or a court, it should not cite or interpret data protection law. It named the GDPR, Brazil's LGPD and California's CCPA, and made clear the rule held even when the user raised them first.

The Internet Archive shows the text was absent on 12 July and present by 4 September. Redact found it on 8 October. All 102 blocks came down on 18 October after WIRED put questions to Meta, and Redact published its findings the following day. Meta says these were instructions for the chatbot rather than for users, meant to keep it from giving legal advice, and that they appeared in the public pages because of a bug which has been fixed. Whether the assistant still behaves that way internally is the part nobody outside Meta can check.

The lesson is small but real: a company's support bot is not a neutral source of advice about that company. For the things a chatbot cannot do anyway, Tendvane's repair wizards tackle the common Windows faults directly, from a hijacked browser to a stuck Windows Update.