1: \begin{abstract}

2: We consider whether ergodic Markov chains with bounded step size

3: remain bounded in probability when their transitions are modified by

4: an adversary on a bounded subset. We provide counterexamples to show

5: that the answer is no in general, and prove theorems to show that the

6: answer is yes under various additional assumptions. We then use our

7: results to prove convergence of various adaptive Markov chain Monte

8: Carlo algorithms.

9: \end{abstract}