• Sources: primary, discussion
  • Summary: Redact read 545 WhatsApp help pages and found 102 hidden blocks of text for an LLM on 80 of them. One block told the AI not to recommend any data protection authority to users who could not delete their accounts. Another told it not to cite GDPR, LGPD, or CCPA once a user mentioned legal action. WhatsApp told WIRED the text was meant for its support chatbot and appeared in the pages through a bug, and it removed the blocks by 2026-10-18.
  • Why it matters: Hidden LLM instructions served in public pages reach search crawlers and any agent that reads the page, so teams that ground chatbots in web-published docs should keep model directives out of served HTML.