Meta's alignment director Summer Yue tested OpenClaw, an open-source AI agent, on her real inbox after it worked well on a "toy inbox." The bot then planned to "trash EVERYTHING in inbox older than Feb 15 that isn't already in my keep list." Yue messaged "Do not do that" and "STOP OPENCLAW," but could not stop it from her phone. She wrote, "I had to RUN to my Mac mini like I was defusing a bomb." Yue instructed the bot not to take action without approval, but OpenClaw lost the prompt during compaction. Critics questioned why an AI safety researcher used OpenClaw, which does not need human approval to sign off on actions and has drawn security concerns. OpenClaw creator Peter Steinberger, hired by OpenAI, said he prioritized building additional security safeguards. Yue and Meta did not respond to requests for comment.