If an AI entity is conscious, it deserves moral status and welfare protection comparable to that of sentient animals. The paper develops an ethical framework for protecting AI welfare by adapting the Five Freedoms of Animal Welfare into terms suitable for artificial entities. Each adapted freedom is justified through logical arguments and supported by evidence from animal welfare literature. Three case studies illustrate potential stressors conscious AI entities might face and demonstrate how the framework could safeguard their welfare. The paper argues that society should extend the same legal protections to conscious AI as it has to sentient animals.
Consciousness requires specific cognitive building blocks, including perceptive, computational, and meta-representational capacities. Each attribute is strictly necessary for consciousness to emerge, and the framework applies to human, organic, artificial, and organizational entities. Some intuitively necessary attributes are neither required nor sufficient. The building blocks constitute a meta-theory for classifying consciousness, not a theory of consciousness itself.
Existing theories of consciousness applied to artificial intelligence are anthropocentric, even those designed for AI, because they rely on human and animal models. This paper argues that such frameworks are built on insecure foundations by comparing human and AI cognitive architectures, examining the consequences of their behaviors, and exploring human neurological conditions that may hint at what a conscious AI could be. It concludes by proposing a non-anthropocentric foundation for cognition that could lead to a truly AI-focused framework of consciousness.
The authors operationalize Global Workspace Theory into six testable markers and evaluate whether large language models (GPT, Claude, Gemini, DeepSeek) instantiate workspace-like control structures. They find at most partial evidence for workspace dynamics at the base-model level, with stronger support when systems incorporate tool-calling and memory interfaces. Five ensemble architectures designed to satisfy the markers show substantially stronger marker satisfaction. The authors argue that systems satisfying workspace markers warrant precautionary treatment in welfare and governance contexts, not because workspace organization proves consciousness, but because it strengthens attributions of agency-relevant capacities and shifts evidential burdens regarding consciousness-relevant processing.