Recent mishaps involving OpenClaw and other AI assistants highlight why entrusting too much authority to automated AI agents is a serious mistake