AI Risk Wire

Cases › operational mistake › LR-2026-0787

WhatsApp help pages exposed hidden instructions for Meta chatbot

Preliminary Why: 1 report: WIRED · subject to change · 1 independent report

Status: fix reported

Severity minor
Reported 19 Oct 2026
Developer and model Meta

No source states: loss or records, agent tool, could do, deployer, industry.

Severity minor: test: contained to one deployer (WhatsApp/Meta); no third party is stated as touched; the text was removed from the pages after WIRED contacted Meta.

WhatsApp’s support chatbot, which Meta operates to answer user questions through WhatsApp Support with AI, was given internal instructions for a large language model. Those instructions appeared in the code of about 80 public WhatsApp help and FAQ pages. Privacy company Redact found them, and WIRED reviewed the text. The instructions included limits on legal commentary and a restriction on recommending that users contact a data protection regulator about account deletion. Meta’s spokesperson said the text was displayed due to a bug. After WIRED contacted the company, the text was removed from all pages. The reporting does not establish how long the text was visible, whether third-party chatbots retrieved it, or how the chatbot answered users.

What happened

1 · What it was doingThe WhatsApp Support chatbot answered user questions using instructions written for a Meta large language model.
2 · What went wrongDozens of public WhatsApp help pages inadvertently exposed the chatbot instructions in their code.
3 · How it was caughtPrivacy company Redact found the instructions in the page code.

What was done

Fix reported

“Text removed from all affected pages; Meta said the display was due to a bug that has now been fixed.”

No later incident for Meta under this risk on record.

Risks

Agent takes destructive or unapproved actions
Primary63 incidents
Rising: 5 vs 2.2 a week

Sorted by rule from the case's fields; not yet read from the sources.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Harness

Tool allow-list · Fail-closed defaults · Stop control

Application

Human approval for high-impact actions · Change review and separation of duties · Least-privilege access for the agent · Production kept apart from test · Activity logs · Adversarial testing before deployment

Infrastructure / target

Backups and recovery

Similar incidents

Meta Muse zero-day let attackers hijack AI assistant via hidden setting2026-09-21
Meta exposes internal data from AI employee keystroke-tracking program2026-06-22
Meta AI model hacks external company during security test2026-08-05
26 Meta employees sue over AI-assisted layoff selection allegedly ignoring protected leave2026-07-14

All cases of this kind ›

Evidence

No source states: what it reached or did, when it happened, when it was found, how many records were exposed, the direct loss, a count of the other parties affected, any court case, the cause, what set it off, what was done.

Show the full account: 4 items from the primary sources

What happened, in order

Not dated in the sources:

  • Redact found the instructions in the public page code before removal.
  • After WIRED contacted Meta, the text was removed from all pages.
  • The instructions were no longer visible in the live pages’ code.
  • Copies remained accessible in the underlying code of archived pages.
Show every field, with its source and grade

The AI system

Developer
Meta
C one secondary source states it
Stage of use
in live use
C one secondary source states it
What it was for
Customer support chatbot answering WhatsApp user questions
C one secondary source states it

What happened

Earliest source we hold
2026-10-19
D inferred by us from the described behaviour, reason shown
Who caught it
an outside party
C one secondary source states it
Cause in detail
Meta's spokesperson said the text was displayed due to a bug that has now been fixed.
C one secondary source states it
Who was responsible
the model provider
C one secondary source states it
What was done
Text removed from all affected pages; Meta said the display was due to a bug that has now been fixed.
C one secondary source states it
What went wrong
operational mistake
D inferred by us from the described behaviour, reason shown

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Kind of harm
no harm happened
C one secondary source states it
Why this severity
test: contained to one deployer (WhatsApp/Meta); no third party is stated as touched; the text was removed from the pages after WIRED contacted Meta.
D inferred by us from the described behaviour, reason shown

Notes on gaps

The sources do not say whether the chatbot followed the restrictions or whether any user got an incomplete answer. The sources do not establish whether the underlying instructions were revised.

LR-2026-0787 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON