Significant economic losses to communities have been incurred even when buildings are designed in accordance with code provisions and local construction practices of the time. This is due to the fact that codes have emphasized on ensuring life safety of the buildings, but not the socio-economic impact on communities from the damage or loss of functionality of the buildings. Additionally, in the development of those codes the building’s interaction with the community was not considered and the consequence of building damages and failures was determined in isolation. As a step forward to improve seismic design practice, the target safety levels that the design standards are based on should be properly engineered by considering community-level objectives. This can lead to appropriate management of the seismic risk that a community is faced with. Therefore, this study aims at developing a framework for optimizing the target safety levels of a design standard based on community-level objectives such as minimizing the regional loss, minimizing the recovery time for a given time horizon. In the evaluation of community-level objectives, the proposed framework can incorporate any functional interdependencies between different buildings in a community. The optimization process requires structural analyses of multiple possible designs of different buildings in a community. If done for good accuracy, the analysis is computationally demanding and makes it difficult to implement the optimization process in the determination of the target safety levels. Instead of such time-consuming structural analyses, the framework utilizes neural networks for estimating the responses. To illustrate the framework, optimization of target reliability indices reflected in the importance factors of seismic design load standards is performed.