Ductile damage in dual-phase steels deteriorates the mechanical properties, resulting in a lowered crashworthiness and a shorter life-time under cyclic loading. However, most works focus on the observation of damage under simple strain paths, even though work pieces are often subjected to combined strain paths during forming. In this work, we set out to characterise the damage behaviour for a DP800 steel subjected to combinations of tensile and bending loads. To this end, we use an automated approach based on machine learning to quantify damage in high-resolution SEM panoramic images, coupled with finite element modelling to determine the stress state, and high-throughput nanoindentation to determine the strain hardening behaviour. We find a strong connection between loading direction and damage formation in both steps of the forming process. By focussing on different possible combinations of positive and negative triaxiality, we show that the combination of two deformation modes with positive triaxiality leads to a promotion of damage formation. A negative triaxiality applied after a positive triaxiality leads to void closure, reducing the number of damage sites and total void area. Importantly, deformation with negative triaxiality carried out before deformation with positive triaxiality inhibits damage nucleation and growth due to strain hardening.