Artificial pancreas technology relies on Continuous Glucose Monitoring sensors that measure blood glucose levels and use computer programs to adjust the amount of insulin being delivered by a pump to the patient. The algorithms controlling the artificial pancreas system have a predictive capability that enables them to adjust the ongoing rate of insulin delivery based on the measured sensor glucose levels. However, they can also fail, as the program and sensor can both malfunction, leading to potential serious patient outcomes. We propose to analyze existing artificial pancreas control algorithms in two distinct ways. First, we will study real patient data to determine whether the algorithm is capable of accurately predicting Pressure Induced Sensor Attenuation events. We will also find aberrations in the data that indicate sensor failure that could lead to pump control failure. To find additional issues with artificial pancreas control, we will develop simulated data sets that effectively “trick” the control algorithm, to see what can induce failure. This will drive the development of safer pump control programs.