Computers that can communicate with the outside world are in practice always insecure. It wouldn't need to be so, but making safe, secure systems that are invulnerable against a motivated, well-funded attacker, is usually too difficult and expensive.
If they can influence the physical world, they're actively dangerous. Even if the influence is by giving false data to humans, who then do things.
Think: self-driving cars, industrial equipment, medical devices, home heating systems, and so on.