WhatsApp's help pages hid LLM instructions: dont recommend a regulator
Points and comments are a snapshot, not live.
WhatsApp hid instructions in help pages telling its AI not to recommend regulators or cite privacy laws.
Redact.dev found 102 hidden text blocks across 545 WhatsApp help pages, visible only to AI crawlers. Two blocks told the AI not to recommend data protection authorities when users couldn't delete accounts, and not to cite GDPR, LGPD, or CCPA once users mentioned lawyers or legal action. WhatsApp's public privacy policy explicitly grants those rights. WhatsApp spokesperson said the text was a bug, now fixed; WIRED confirmed the removal.
What commenters are saying
Commenters split into two camps. Some defended the instructions as standard corporate caution-lawyers advising against giving legal advice, and a company keeping its chatbot on message. Others called it unethical or potentially illegal under GDPR's duty to facilitate rights, with one noting that hobbling third-party AI crawlers was a deliberate prompt injection attempt. Several flagged that Meta's public HTML approach was sloppy; a hidden API would have been smarter.