After the discovery of the Higgs boson at the Large Hadron Collider at CERN, a new time period characterized by a lack of discoveries of Beyond Standard Model physics in particles accelerators is ongoing. Anomaly detection is a novel machine learning approach that can represent a new possible approach to searches, as it allows a very general method with the signatures of interest without losing sensibility to possible signals. ATLAS analyses are taking the first steps in this direction, following the results obtained from Classification Without Labels, often referred to as CWoLa, based resonant searches. These proceedings show the results obtained with anomaly detection approaches in ATLAS, where events are selected solely because of their incompatibility with a learned background-only model. In particular, the focus is on the search for a heavy resonance 𝑌 decaying into a Standard Model Higgs boson 𝐻 and a new particle 𝑋 in a fully hadronic final state, which represents the first application of fully unsupervised machine learning in an ATLAS analysis.