We consider two different models for adversarial errors and show how to design circuits that can recover from such errors. 

Change of measure techniques are ubiquitous in many areas of applied probability including in model selection in finance and in efficient simulation of rare events.