Attempts at hijacking AI resources are now taking place on an industrial scale. How is AI infrastructure being targeted, and what defensive measures should you implement?