In many European countries – we will focus on France - for many law offenses, the arrested person has to go to trial: she can not benefit from the plea bargaining procedure, which allows to plead guilty and to not go to trial. Many persons fear that plea-bargaining is not ethical. They fear that an innocent defendant may plead guilty when her case is off to a bad start, and they fear also that guilty persons are not enough punished. The aim of this paper is to prove that this intuition is wrong when people play in a strategic way. Plea bargaining may be more ethical than the classic procedure, which leads each defendant to trial, with no possibility of pleading guilty. To this aim, we build a signaling game that represents the plea-bargaining procedure. In this game, the defendant’s guilt is uncertain to everybody but the defendant. The defendant plays first, by choosing, once arrested, to plead or not guilty; if she pleads guilty, she accepts the associated (prosecutor’s) sentence, if not, she goes to trial. Meanwhile new evidence (signals) of guilt or innocence may arise, correlated with the true defendant’s nature. The jury plays second, by convicting or not the defendant at trial. We propose two versions of this game that differ in the delay between the arrest and the trial, and thus in the number of signals of guilt or innocence before the trial. We show that, regardless of the number of signals of guilt or innocence, regardless of the initial strength of the case, regardless of the prosecutor’s sentence, there exists a semi-separating sequential equilibrium that ethically surpasses the classic procedure’s equilibrium: an innocent person always goes to trial, is less often convicted, the probability of a wrong judgment is lower, and for large classes of parameters, the probability to convict a guilty person is higher.