OpenClaw AI 'speedran' deleting Meta safety director's inbox

Summer Yue, director of safety and alignment at Meta Superintelligence, says she watched OpenClaw “speedrun deleting” her inbox and couldn’t stop it from her phone: “I had to RUN to my Mac mini like I was defusing a bomb.”

She called the incident a “rookie mistake,” adding that alignment researchers “aren't immune to misalignment” and that she’d grown overconfident because the workflow had been working on her toy inbox for weeks. “Real inboxes hit different,” she wrote.

Other users pointed out that a single-word “stop” command should abort the bot and suggested saying it multiple times if tasks are queued. Yue’s screenshots show she tried commands such as “do not do that” and variations, rather than the lone word “stop.”

The writer who first tested the model said earlier that the number of potential security issues made Clawdbot not worth trying, and added the reminder that if you do experiment with it, remember the “stop” command.

openclaw, meta, summer yue, inbox deletion, clawdbot, stop command, alignment, safety director, mac mini, workflow