The Chinese wall model combines discretionary and mandatory aspects of access control. Hence it is widely used in commercial environments to prevent information flows from competing companies with conflicting of interests to the same consultant. However, the model gives strong constraints on both reads and writes, so it is too restrictive to be employed in a practical system. Especially for data leakage prevention (DLP), the applications not play to its advantages. This paper reconsiders the conflict of interest from the perspective of the data object and put forward the concept of aggressive conflict of interest relation. The new relation extends the constraints on two-way information flow to that of one-way flows. Based on this, the paper presents an aggressive Chinese wall model (ACWM) for initiative data leakage prevention and gives the formal description of the model as well as the related proof of the theorem. The final analysis shows that, ACWM achieves the same security goal as traditional Chinese wall models, and also provides more flexible constraints which are efficient for DLP.