Despite hundreds of searches, no clear evidence of beyond standard model particles have been found at the LHC so far. However, there still remains a huge landscape of possible signals hiding in LHC data, which is very difficult to fully cover with traditional strategies. Recently, a new class of ‘anomaly detection’ techniques have been developed which aim to cover as much of this landscape as possible so that potential discoveries are not missed. These techniques leverage novel data-driven AI methods to minimize the trade-off between sensitivity and model independence. In this talk, I will present results from the first application of anomaly detection in CMS : a search for resonances decaying to two jets with 'anomalous' substructure. I will also discuss future prospects for these exciting new AI techniques more broadly, and how they may play a key role in a future, robust LHC search program.
