This paper presents a model for selecting and prioritizing in environments of uncertainty, a set of critical threats in a software development project and so, to minimize risks and to increase the level of performance from the user's requirements. The proposed model is based on the principles of artificial intelligence technique "Rough Set" and allows to obtain, to spread and to use the acquired knowledge by the team of work throughout the project characterizing, classifying and reducing in a recurrent and incremental way the possible threats to which the project may be exposed. The model was validated with state agencies projects in the city showing a progressive improvement of 16.48% in the performance from the user's requirements.