Full likelihoods encode the entire statistical model of an analysis and thus range among the most invaluable analysis data products for a large range of analyses, ranging from SM measurements to BSM searches. ATLAS has recently started to release the first full analysis likelihoods using a python-based implementation of HistFactory. In this talk, the JSON specification used to release the likelihoods in serialisable format is discussed and details on how process them are given.
In addition, a tool to build simplified likelihoods targeted for CPU-intensive large-scale reinterpretations is presented.
Finally, the current collaboration policy and future plans are discussed.