WhatsApp’s Hidden Notes Told AI Not to Recommend a Regulator
An investigation found hidden AI-directed notes on WhatsApp help pages that told a chatbot not to recommend data-protection regulators and, in some legal contexts, not to cite privacy laws. WhatsApp removed the notes on October 18; it told WIRED they were for its own support bot and appeared on the pages because of a bug. The report does not establish that outside AI assistants followed them.
Redact says it reviewed 545 linked help pages and found 102 hidden blocks across 80 pages. It highlights the conflict between visible privacy-policy disclosures and hidden instructions that could steer support answers, while also acknowledging the legitimate concern that a company chatbot should not give legal advice. The author says the live pages no longer contain the notes; archived copies preserve examples.
HN readers sharply disagreed: some saw the regulator and privacy-law restrictions as withholding useful information, while others defended guardrails against a support bot making unreliable legal claims. Commenters also disputed whether the hidden-page mechanism could affect third-party chatbots and whether WhatsApp’s explanation of a display bug was credible.