Higher-spin gravity is a curious beast of mathematical physics: a cousin of supergravity and string theory that seems comfortable with 4 spacetime dimensions and positive cosmological constant. On the other hand, general arguments show that this theory must be pathologically non-local at quartic order. In this talk, I claim that the non-locality arguments rely on Lorentzian boundary signature. For Euclidean boundary, explicit calculation shows that the feared non-locality is absent. This implies that the theory is healthy in de Sitter space, but not in (Lorentzian) AdS. The surprising possibility of such signature-dependent locality has long been implicit in the CFT/holography literature. Higher-spin gravity provides the first explicit example.