“Paul views the security vulnerabilities exposed by OpenClaw and MoltBook as a prime example of the risks of allowing uncontrolled "Shadow IT" AI agents within an organization.”