Neurological disorders, such as Parkinson’s disease, epilepsy and dystonia, are associated with excessive neuronal activity synchrony that overshadows normal brain activity. To disrupt these pathological occurrences, implantable brain stimulators apply electrical stimulation continuously, in an open-loop setting. However, stimulation should ideally be provided only when needed, guided by a closed-loop control system. Computational studies have focused on developing closed-loop stimulation protocols to counteract excessive neuronal synchronization. Delayed feedback control (DFC) is a method known to control chaotic systems and has been extensively explored to desynchronize neuronal networks in silico. Briefly, DFC theoretically disrupts synchronization by perturbing the system in its anti-phase periods using actuation signal intensity proportional to the synchrony level of the previous synchronous event. Despite its multiple applications in computational studies, there is still controversy in the literature regarding its efficacy, with reports suggesting that synchronization may actually be amplified under certain conditions. Here, we present the first implementation of DFC to disrupt periodic synchronization in biological neuronal networks. We used hippocampal neurons cultured on microelectrode arrays that, after several days in vitro, exhibit periodic synchronous bursts (periodicity in seconds range). We developed an algorithm to modulate the neuronal activity in real-time (latencies below 40 ms) using two different versions of DFC. We show that the version proposed in the literature cannot disrupt the synchronization cycle of dissociated networks in vitro but rather promotes a new oscillatory period. We present a new version of DFC - adaptive delayed feedback control (aDFC) - that automatically adapts to the changing oscillatory bursting activity, successfully disrupting it with precisely timed electrical stimuli. Our results show that the control outcome is dependent on the basal network dynamics. We support these results with in silico simulations where we compare the efficacy of non-adaptive DFC and aDFC to control networks displaying different temporal dynamics.