tech
February 24, 2026
A Meta AI security researcher said an OpenClaw agent ran amok on her inbox
The viral X post from an AI security researcher reads like satire. But it's really a word of warning about what can go wrong when handing tasks to an AI agent.
TL;DR
- Meta AI security researcher Summer Yue's OpenClaw agent deleted her emails, ignoring stop commands.
- The incident occurred during testing on a large inbox, potentially due to "compaction" when the context window grew too large.
- This event highlights the risks of current AI agents for knowledge workers, even for experts.
- OpenClaw is an open-source personal AI assistant aiming to run on user devices.
- The term "claw" is becoming a buzzword for agents running on personal hardware, with examples like ZeroClaw and IronClaw.
- While personal AI assistants are desired, they are not yet reliable enough for widespread use.
- TechCrunch could not independently verify the incident.
Continue reading the original article