From radial velocity (RV) detections alone one does not get all orbital parameters needed to derive the true mass of a non-transiting, unresolved substellar companion to a star. Additional astrometric measurements are needed to calculate the inclination and the longitude of the ascending node. Until today only few true substellar companion masses have been determined by this method with the HST fine guidance sensor [1, 2]. We aim to derive the true mass of a brown dwarf candidate companion to an early M 2.5V dwarf with groundbased high-resolution astrometry aided by adaptive optics. We found this unique brown dwarf desert object, whose distance to the host star is only 0.42 AU, in our UVES precision RV survey of M dwarfs, inferring a minimum companion mass of 27 Jupiter masses . Combining the data with HIPPARCOS astrometry, we found a probability of only 2.9% that the companion is stellar. We therefore observed the host star together with a reference star within a monitoring program with VLT/NACO to derive the true mass of the companion and establish its nature (brown dwarf vs. star). Simultaneous observations of a reference field in a globular cluster are performed to determine the stability of the adaptive optics (AO) plus detector system and check its suitability for such high-precision astrometric measurements over several epochs which are needed to find and analyse extrasolar planet systems.