AI Hallucinations: Is “Artificial Evil” Possible?

The spread of digital technologies, especially AI systems, in various areas of modern society is provoking a large number of ethical risks and related problems. In the field of theoretical ethics, this has led to numerous attempts to critically rethink a number of moral and philosophical concepts, including such fundamental categories as good and evil. One consequence of this: the emergence of the idea that, along with such traditionally studied concepts as natural evil and moral evil, the use of AI system technologies presupposes the existence of an artificial evil. This idea has become all the more relevant in connection with the widespread use of GPT-based technologies, since, from the point of view of mass public consciousness, such AI is often perceived as comparable to human intelligence. This perception is reinforced by the AI’s reoccurring hallucinations, similar to the ability to lie. This article is devoted to research on the specifics of moral evil, a critical analysis of the hypothesis about the existence of AE, AI hallucinations in their moral meaning. As one of the possible ways to prevent AI hallucinations or mitigate ethical risks associated with their consequences, the hypothesis of the creation of an AI ethical Supervisor is considered, including its possible foundations such as “negative ethics”. The main conclusion is that currently there are no digital technologies, including AI systems, that could be considered as a possible AE. In this regard, preventing ethical risks and moral evil is the responsibility of people developing and using digital technologies.