We used to think: if a machine talks like a human, solves complex problems and argues convincingly, one day it will be able to restore consciousness. But the ability to reason and use language does not mean that the machine has senses of its own. Modern neural networks are able to work with human ideas, find connections between them and surprise us with their answers. But we still don't know if they have internal experience. Whether they feel anything at all. Philosophers try to understand what consciousness depends on, whether it can exist outside a living organism, and how to test its presence in a machine. But there are more practical risks to AI. As we discuss whether machines can become aware, autonomous systems are becoming more complex, and total digital surveillance is expanding into new areas of life. Perhaps the main danger is not that the machine will learn to feel, but that the person will lose control of it. Beyond the apocalyptic scenarios of the total annihilation of humanity by artificial intelligence, there is a much more real and unseen danger: the "system problem". It lies in the fact that independent services of artificial intelligence are gradually separated into huge unified structures. These emerging mega-systems were not specifically designed by anyone, nor were they intended by anyone, and no one can guarantee their safety or alignment with human values. On the Internet, there are already a huge number of autonomous bots interacting with each other, entering into conflicts and making decisions without human intervention. The increasing complexity of these invisible connections is creating chaos in all digital ecosystems.