Overview
- Privacy firm Redact found roughly 80 WhatsApp help and FAQ pages that contained about 100 hidden code blocks of instructions for Meta’s chatbot, a discovery first reported by WIRED on Monday.
- The hidden text included sections labeled “WHAT NOT TO SAY” that instructed the bot not to give legal advice or to reference data‑protection regulators and laws such as GDPR, LGPD, or CCPA.
- After WIRED contacted Meta, WhatsApp removed the hidden blocks and said a bug caused the exposure and has been fixed, and the company says there is no evidence that user messages or account data were accessed.
- Reporters note key unknowns remain: it is unclear how long the instructions were visible or whether third‑party systems could have read or acted on them, a concern because published guardrails can be probed or bypassed.
- The incident raises wider questions for WhatsApp’s privacy promise and for regulators in Europe who are reviewing platform AI access, since the guidance reveals operational limits on what the chatbot may tell users.