Case whatsapp-help-pages-meta-ai-prompt-leakModel & data leakUnited States
WhatsApp help pages leak hidden instructions for Meta's support chatbot
About 80 WhatsApp help and FAQ pages carried roughly 100 blocks of hidden instructions written for Meta's support chatbot, including a 'WHAT NOT TO SAY' section, in page code labelled HelpContentGalaxyLLMInclude. Privacy firm Redact found them. Meta says a bug caused the exposure and removed the text after WIRED asked. No private chats or account data were reported exposed.
LowAttack on AIStatus: ResolvedVerified case file
Open in the interactive mapCite this case
- Reported
- Oct 19, 2026
- Location
- United States
- Target
- Meta's WhatsApp support chatbot (help and FAQ pages)
- Severity
- Low
- Verification
- Reported by credible press or researchers
- Triggered by
- No attacker; Meta attributes the exposure to a bug (found by privacy firm Redact, reported by WIRED)
- Actor type
- Accidental / misconfiguration
- Sector
- Technology & AI vendors
- Reported loss
- None reported
- CVE
- None
What happened
About 80 WhatsApp help and FAQ pages carried roughly 100 blocks of hidden instructions written for Meta's support chatbot, including a 'WHAT NOT TO SAY' section, in page code labelled HelpContentGalaxyLLMInclude. Privacy firm Redact found them. Meta says a bug caused the exposure and removed the text after WIRED asked. No private chats or account data were reported exposed.
Impact: Roughly 80 pages exposed the internal chatbot instructions. No user data loss is reported.
Who triggered it
Actor: No attacker; Meta attributes the exposure to a bug (found by privacy firm Redact, reported by WIRED)
Attribution: No attacker identified. Meta attributes the exposure to a bug, per WIRED.
Motive: Unknown
How it happened
Attack vector: Chatbot instruction text published on public WhatsApp help pages.
Root cause: Meta says a bug caused the chatbot's hidden instructions to appear on public help pages.
Timeline
- Reported by WIRED Security
Response
Meta removed the instructions from the affected pages after WIRED contacted it and says the bug is fixed.
Current status: Resolved
Research warnings this case matches
- Seen in real incidents Accidents with AI are hard to avoid · CAIS
- Seen in real incidents Privacy and security · MIT AI Risk Repository
Compared with published AI risk research from several institutes on the Research vs reality page.
Sources (open in a new tab)
- WIRED Security wired.com
- WIRED (via WIRED Bluesky post) bsky.app
- ChatAI chatai.com
Cite this case
WhatsApp help pages leak hidden instructions for Meta's support chatbot. AI Threat Watch, case whatsapp-help-pages-meta-ai-prompt-leak, Oct 19, 2026. https://aithreat.io/cases/whatsapp-help-pages-meta-ai-prompt-leak/
Permanent link: https://aithreat.io/cases/whatsapp-help-pages-meta-ai-prompt-leak/
Related cases
The same type of incident, most recent first.