Lech Mazur comments on Sabotage Evaluations for Frontier Models