Demands for public participation in technical decision-making are currently high on the agenda of Science & Technology Studies. It is assumed that the democratisation of technical decision-making processes generally leads to more socially desirable and acceptable outcomes. While this may be true in certain cases, this assumption cannot be generalised. I will discuss the case of the so-called ‘South African AZT debate’. The controversy started when President Thabo Mbeki, after reading some scientific papers on the toxicity of AZT, decided to (...) bar the use of the drug in the public health sector as a means to reduce the transmission of HIV from mothers to children. While the scientific mainstream accepts the effectiveness of AZT in reducing the risk of vertical HIV transmission, a few maverick scientists reject the clinical evidence and argue that the risks of using AZT by far outweigh its benefits. Based on various textual sources and using the ‘Periodic Table of Expertises’ developed by Collins and Evans, Mbeki’s expertise at the time of his intervention into the technical question whether AZT is a medicine or a poison can be classified as primary source knowledge. It is shown that this type of expertise is insufficient for technical decision-making. Mbeki’s primary source knowledge legitimated his presentation of the claims of maverick scientists as a serious contribution to the debate—with tragic consequences for tens of thousands of babies.Keywords: Primary source knowledge; Expertise; Technical decision-making; South Africa; AZT; AIDS. (shrink)