Assessing risks in the petrochemical industry involves understanding the potential behaviour of accidentally released clouds of various substances. Anhydrous hydrogen fluoride (HF) poses a particular problem. On the one hand its chemical reactivity is a severe hazard, and on the other, its complex thermodynamic behaviour means that it is difficult to model accidental releases in order to estimate the hazard. In this paper we examine the expected behaviour of hydrogen fluoride clouds using the code DRIFT, and contrast it with that of substances which are easier to model. A comparison is given between results from DRIFT and from the Goldfish HF field trials. Furthermore, we show that the homogeneous equilibrium model for the thermodynamic properties of the cloud can describe quite complicated phenomena and that atmospheric humidity, while less important for other substances, may be an important factor in hydrogen fluoride dispersion. This is illustrated by examining the effects that atmospheric humidity may have had on the Goldfish trials had they been conducted in less arid circumstances.