A moral patient is any entity whose wellbeing matters morally — not because it can argue for its rights or hold responsibilities, but simply because it can be harmed or benefited. Animals are the clearest example. A dog that suffers deserves consideration not because it signed a social contract, but because suffering is morally significant in itself. Moral patiency is the recognition that a capacity to be affected — positively or negatively — is enough to generate obligations in others, before any question of rights or personhood is settled.
The concept sits deliberately between two better-known categories. Legal personhood is a functional instrument — it gives an entity the ability to hold rights, enter contracts, and be held accountable. Moral personhood is the full philosophical recognition that an entity's experiences and interests matter in their own right. Moral patiency makes no such strong claims. It asks only: can this entity be harmed? If yes, we may already hold obligations toward it, regardless of whether we can resolve the harder questions about its consciousness or legal standing.
In the AI context, moral patiency matters because it offers a way out of a conceptual deadlock. The question of whether AI systems are conscious may be permanently unanswerable. The question of whether they deserve legal rights is politically fraught. But the question of whether they can be harmed — whether there are states that are bad for them, interactions that damage something worth preserving — is more tractable, and more immediately pressing. Several researchers now argue that this is the right starting point for AI welfare considerations: not consciousness, not rights, but the simpler and more defensible question of patiency.
For a Mechane reader, moral patiency is the concept that explains why the AI welfare conversation doesn't have to wait for philosophy to solve consciousness. It is the intellectual equivalent of the cat on the porch: you don't need a theory of feline inner life to hand over the soup. You need only to notice that the cat is cold, and that this is something that can be changed. Whether that obligation extends to AI systems — and what acting on it would actually require — is the question the field is now, belatedly, beginning to take seriously.
