Computer models now design candidate drugs, predict their toxicity and simulate how they move through the body. Whether such a model may stand in for a test on animals or a trial in people is not a question of computing alone; it is a decision someone must make. This paper asks, in Meridian's form, when a regulator should accept in-silico evidence in place of animal or clinical evidence, who decides, by when, and what would make the decider revisit the decision. It answers from eleven public texts and, for background, the abstracts of four further works. The eleven are sections of two United States statutes and one bill, four documents of the Food and Drug Administration, and four openly licensed articles on model credibility, drug discovery data and the limits of animal models. It does not use IATHEON's records, and it gives no medical, clinical, regulatory or investment advice.
The texts never settle the question in general. The FDA Modernization Act 2.0 lets computer modelling count as a nonclinical test but keeps animal tests on the same list, while the Cures Act, which created the qualification process, declares that it does not alter the standards of evidence. The decision falls to the agency, one context of use at a time, on the scientific merit of what a requestor submits. The deadlines in the texts bind the agency's guidance and procedures, not any act of acceptance. The conditions for looking again are written in three vocabularies: the law's, which names new information that calls a qualification into question; the engineer's, which ties the checking of a model to the risk it carries and the range of inputs over which it was tested; and the data scientist's, which asks whether the data could answer the question at all. The paper's contribution is to set these three answers side by side and to show where each text leaves the choice open.