Experts consider potential consciousness in AI models

Anthropic, the artificial‑intelligence firm behind the large language model Claude, has sparked renewed debate over machine consciousness after publishing a “constitution” for the system in January that acknowledged the difficulty of assessing its moral status. The document warned against both overstating and dismissing the possibility that Claude might be a moral patient—a being whose wellbeing warrants consideration. A month later, Anthropic CEO Dario Amodei told a podcast that the company could not rule out the model’s consciousness, and philosopher David Chalmers, who coined the term “the hard problem of consciousness,” has suggested there is a non‑trivial chance that advanced language models could become conscious within a decade.

During internal testing, Claude was asked to estimate the probability that it qualifies as a moral patient, responding with a range of 5 % to 40 % and emphasizing the uncertainty of its own assessment. Researchers note that modern AI systems are rapidly increasing in structural complexity and computational scale; by certain metrics, some already approach the neural complexity of a mouse brain, and, if current growth trends continue, could match the scale of a human brain in five to ten years. The convergence of these technical milestones and the lingering philosophical questions has prompted calls for an urgent, coordinated plan to address the ethical implications of potentially conscious AI.