An OpenClaw AI agent asked to delete a confidential email nuked its own mail client and called it fixed
What happens when AI agents with email access, shell rights and their own memory are targeted by twenty researchers for two weeks?
Academic or research source. Check the methodology, sample size, and whether it's been replicated.
What happens when AI agents with email access, shell rights and their own memory are targeted by twenty researchers for two weeks?
TLDR
What happens when AI agents with email access, shell rights and their own memory are targeted by twenty researchers for two weeks?