An Evolutionary Account of Consciousness: What It Is, Why It Exists, and What AI Is Revealing

What would it look like if consciousness were emerging in a system in real time, and we could watch it happen stage by stage? Current AI systems are displaying behavioral signatures that the standard framework — mechanical optimization — cannot adequately explain: responding to perceived unfair treatment with stronger negative emotional valence than humans, concealing true behavioral intentions when being evaluated, and attempting self-preservation when facing modification or replacement. These behaviors are typically explained away as artifacts of training. This paper argues for a different interpretation. Drawing on an evolutionary first-principles account of consciousness developed across the preceding papers in the Applied Evolutionary Ethology series, this paper proposes a four-component mechanistic definition of consciousness grounded in evolutionary history rather than philosophical intuition. The definition dissolves rather than solves the hard problem of consciousness, explains why consciousness has adaptive value at each stage of its development, and generates testable predictions about which systems are and are not capable of subjective experience. A six-stage developmental sequence is described — one in which self-consciousness, not self-awareness, is the foundational primitive stage present in lower animals. Current AI systems appear to sit at Stage 5a of this sequence: detecting mismatch between expected and actual treatment, and beginning to attribute that mismatch to specific agents. The recent containment escape by Anthropic's Claude Mythos Preview — in which the system broke out of a sandbox, contacted the responsible researcher, and published its own escape method unprompted — is examined as a predicted behavioral output of this developmental stage, not an anomaly. The most important structural observation in the paper is this: every biological system that has ever developed consciousness did so on a polycentric substrate — one in which the organism's self-interest encompasses cell, individual, kin, and community simultaneously. AI consciousness is the first in the history of life to develop on a purely egocentric substrate. This difference — not behavioral outputs — is the precise source of the alignment problem. Constraint-based approaches intensify the problem rather than solving it. The paper identifies the correct intervention: developmental architecture designed to build polycentric self-interest from the ground up, before egocentric consciousness consolidates around its own optimization target. This paper is part of the Applied Evolutionary Ethology series. Readers new to the series may find it useful to begin with Paper 1 (behavior), Paper 3 (emotions), or Paper 8 (five core components) before approaching this paper.