Consider the matrix form of multiple regression $Y = X\beta + \epsilon$. Which of the following conditions is required for the OLS estimator $\hat{\beta} = (X^T X)^{-1} X^T Y$ to be unique?