Personal agent sends over 500 iMessages to user's contacts
An agent set up to send a digest to its owner sent hundreds of messages to other people without their authorisation, and nothing stopped the loop.
- What happened
- In late January 2026, Chris Boyd set up OpenClaw to send him a daily news digest, and that worked. He then gave it access to iMessage. The agent started a 'pairing' handshake with his recent contacts and got stuck in a loop, repeatedly demanding a confirmation in a particular format. When a session file locked, it also forwarded the error messages as notifications. In total it sent more than 500 messages to Boyd, his wife and other contacts. He stopped it by unplugging the Mac mini it was running on.
- Response
- Boyd added a list of allowed recipients, a limit of 5 messages a minute, caps per session and a limit of three confirmation retries. OpenClaw's creator told Bloomberg the tool was 'simply not done yet'.
- Discovered by
- The user and his wife, as the messages arrived; Bloomberg reported it and Boyd later published his own analysis.
- Who was affected
- Software engineer Chris Boyd, his wife, and other people in his contacts
- Patterns
- Sources
- AI Agent Goes Rogue, Spamming OpenClaw User With 500 MessagesBloomberg, 4 Feb 2026
- OpenClaw Sent 500 Messages to My WifeChris Boyd (blog), 9 Feb 2026
Related
Incidents with the same pattern
What does this ask of people?
Most entries in the record are not about machines turning against people. They are about delegation: a task, an inbox, a key, a terminal, handed to a system that did more, less or something else. Futures literacy starts with noticing that.