Only one candidate MSE formula wrongly tacks on an extra $\mu^2$ term

Let $\{X_t\}$ be a stationary time series model given by: $X_t=\mu+\phi X_{t-1}+Z_t, Z_t\sim iid$ normal $(0,\sigma^2)$. Then, which of the following statements is not true if $P_{X_t}(X_{t+k})$ is the forecast of $X_{t+k}$. Given $X_1,X_2\dots X_t$?