In this paper we propose a methodology to accelerate the resolution of the\nso-called "Sorted L-One Penalized Estimation" (SLOPE) problem. Our method\nleverages the concept of "safe screening", well-studied in the literature for\n\\textit{group-separable} sparsity-inducing norms, and aims at identifying the\nzeros in the solution of SLOPE. More specifically, we derive a set of\n\\(\\tfrac{n(n+1)}{2}\\) inequalities for each element of the \\(n\\)-dimensional\nprimal vector and prove that the latter can be safely screened if some subsets\nof these inequalities are verified. We propose moreover an efficient algorithm\nto jointly apply the proposed procedure to all the primal variables. Our\nprocedure has a complexity \\(\\mathcal{O}(n\\log n + LT)\\) where \\(T\\leq n\\) is a\nproblem-dependent constant and \\(L\\) is the number of zeros identified by the\ntests. Numerical experiments confirm that, for a prescribed computational\nbudget, the proposed methodology leads to significant improvements of the\nsolving precision.\n