Abstract We introduce a generalised subgradient for law-invariant closed convex risk measures on L 1 and establish its relationship with optimal risk allocations and equilibria. Our main result gives sufficient conditions ensuring a non-empty generalised subgradient.