Toxicological hazard assessment currently finds itself at a crossroads where the existing classical test paradigm is challenged by a host of innovative approaches. Animal study protocols are being enhanced for additional parameters and improved for more efficient effect assessment with reduced animal numbers. Whilst existing testing paradigms have generally proven conservative for chemical safety assessment, novel alternative in silico and in vitro approaches and assays are being introduced that begin to elucidate molecular mechanisms of toxicity. Issues such as animal welfare, alternative assay validation, endocrine disruption, and the US-NAS report on toxicity testing in the twenty-first century have provided directionality to these developments. The reductionistic nature of individual alternative assays requires that they be combined in a testing strategy in order to provide a complete picture of the toxicological profile of a compound. One of the challenges of this innovative approach is the combined interpretation of assay results in terms of toxicologically relevant effects. Computational toxicology aims at providing that integration. In order to progress, we need to follow three steps: (1) Learn from past experience in animal studies and human diseases about critical end points and pathways of toxicity. (2) Design alternative assays for essential mechanisms of toxicity. (3) Build an integrative testing strategy tailored to human hazard assessment using a battery of available alternative tests for critical end points that provides optimal in silico and in vitro filters to upgrade toxicological hazard assessment to the mechanistic level.