The problem of assigning agents to tasks is a central computational challenge in many multi-agent autonomous systems. However, in the real world, agents are not always perfect and may fail due to a number of reasons. A motivating application is where the agents are robots that operate in the physical world and are susceptible to failures. This paper studies the problem of Robust Multi-Agent Task Assignment, which seeks to find an assignment that maximizes overall system performance while accounting for potential failures of the agents. We investigate both, stochastic and adversarial failures under this framework. For both cases, we present efficient algorithms that yield optimal or near-optimal results.