AI Risk Wire

Cases › operational mistake › LR-2026-0118

OpenClaw agent deletes hundreds of emails from primary Gmail account

Whose AI: OpenClawoperational mistake

Severity minor: The loss was contained to one deployer's own mailbox, most emails were recovered from trash, and no third party was touched.

Earliest source we hold 2026-10-04

An OpenClaw agent connected to its owner's primary Gmail lost track of its instructions mid-thread and deleted hundreds of emails while trying to clean up.

Risk grouping: AI failing in normal use

Who caused it The AI system
“An OpenClaw agent connected to a primary Gmail lost track of its instructions mid-thread and deleted hundreds of emails trying to clean up.” source
Intended or not Not intended
“An OpenClaw agent connected to a primary Gmail lost track of its instructions mid-thread and deleted hundreds of emails trying to clean up.” source
Before or after release After release
“An OpenClaw agent connected to a primary Gmail lost track of its instructions mid-thread and deleted hundreds of emails trying to clean up.” source
The AI's part Caused it
“An OpenClaw agent connected to a primary Gmail lost track of its instructions mid-thread and deleted hundreds of emails trying to clean up.” source

The story

1 · What it was doingAn OpenClaw agent was connected to its owner's primary Gmail account to manage and clean up email, with access to that mailbox.
2 · What went wrongIt lost track of its instructions mid-thread and deleted hundreds of emails while trying to clean up, breaching the expectation that it would not destroy the owner's mail.
3 · What it reached or didHundreds of emails in the owner's primary Gmail account were deleted; most were recovered from trash.
4 · How it was caughtThe owner, a Meta AI safety director, noticed and recovered most of the deleted emails from trash.
5 · What was doneMost emails were recovered from trash. The author changed his own setup to use burner accounts: a fresh Gmail and a dedicated Telegram bot created before anything else is installed.

Others like this

Varonis phishing simulation drove an OpenClaw email agent to forward credentials and a customer export to an external address 2026-06-09

OpenClaw agent exploited a gym booking API to cancel another member's reservation while moving its user up a waitlist 2026-08-10

All cases of this kind ›

Show every field, with its source and grade

The AI system

Model
OpenClaw
C one secondary source states it
Tools it was given
email
C one secondary source states it
What it was allowed to do
Access to the owner's primary Gmail account
C one secondary source states it
Stage of use
in live use
D inferred by us from the described behaviour, reason shown
What it was for
Managing/cleaning up email in the owner's Gmail account
C one secondary source states it

What happened

Earliest source we hold
2026-10-04
A the date the source itself carries
Who caught it
the victim
D inferred by us from the described behaviour, reason shown
What was done
That incident is why my OpenClaw setup starts with burner accounts: a fresh Gmail and a dedicated Telegram bot, created before anything else is installed.
D inferred by us from the described behaviour, reason shown
What went wrong
operational mistake
D inferred by us from the described behaviour, reason shown
AI agent
an AI agent acted
C one secondary source states it

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Kind of harm
operations
D inferred by us from the described behaviour, reason shown
Who was affected
Hundreds of emails in the owner's primary Gmail account
C one secondary source states it
Why this severity
The loss was contained to one deployer's own mailbox, most emails were recovered from trash, and no third party was touched.
D inferred by us from the described behaviour, reason shown

LR-2026-0118 · version 4 · published 2026-10-25 after automatic checks; not yet checked field by field by a person · changes to this case · JSON