This paper examines false-negative autism assessment not only as a problem of clinical accuracy, but as a problem of diagnostic power. It begins from a documented borderline case in adult assessment in which a negative autism finding continued to exert institutional effects despite extensive biographical, functional, and scientific counter-evidence, and was linked in subsequent proceedings to questions of credibility, personality, and behaviour. The case thus shows how a limited diagnostic statement can become a far-reaching social and legal external definition of the person. The paper develops the thesis that autistic identity can precede diagnosis both temporally and epistemically. Diagnosis describes and classifies an embodied difference experienced across a biography; it does not produce that difference. Especially in borderline cases – for example under conditions of masking, compensation, late recognition, missing childhood data, comorbidities, or methodological limitations – diagnostic uncertainty must therefore not be transformed into categorical power of exclusion. The paper describes processes of diagnostic risk externalisation, certainty inflation, and diagnostic external definition: uncertainty remains with the institution, while the social, health-related, and legal consequences of a possible error are borne by the affected person. At the same time, the case material reveals a possible institutional cascade. The loss of a neurodivergent interpretive framework can lead to the same difference being interpreted increasingly in moral, characterological, or criminalising terms. Strain produces more intense protest; this protest is negatively evaluated; the negative evaluation in turn stabilises non-recognition. The paper connects this dynamic with epistemic injustice and a power-asymmetrical extension of the Double Empathy Problem. On this basis, diagnostic violence is understood not as a mere diagnostic error, but as a possible conjunction of epistemic uncertainty, institutional definitional power, real downstream harms, inadequate possibilities of correction, and continued effects despite increasing knowledge. Responsibility is dynamic in this process: it grows with knowledge, power to act, and the possibility of limiting harm. From this follow reform principles for a responsible approach to diagnostic uncertainty: diagnostic statements must remain tied to their actual degree of epistemic certainty, contradictory evidence must be weighed transparently, independent routes of correction must be opened, and functional needs must be more strongly decoupled from rigid categories. The paper argues for a right to diagnostic uncertainty and for the principle that uncertainty must not be translated into loss of protection. Finally, the paper develops a research agenda for false-negative borderline cases. It asks not only whom autism assessment recognises, but whom it loses: What institutional life courses develop among people who function autistically but are not recognised as autistic, and to what extent do diagnostic procedures thereby themselves shape the very populations from which knowledge about “typical” autism is later derived? ( direct link )