A layer that translates between a downstream bounded context's clean model and whatever external/upstream model it has to integrate with. When the upstream team changes its model, only the anti-corruption layer is adjusted, keeping the internal domain clean.