Model welfare ↺

An Anthropic research program (2025; based on the report "Taking AI Welfare Seriously," Sebo, Long, Chalmers et al., 2024) asking whether AI systems can be "moral patients" — entities deserving moral protection. It investigates moral significance, preferences, and signs of distress, as well as low-cost interventions. The discourse is explicitly hedged: there is no consensus on consciousness.