Safety assessment of infrastructural systems after an earthquake is crucial for informed decision-making regarding emergency actions, rehabilitation, strengthening, and reconstruction, as well as prioritization in budget allocation for post-earthquake reconstruction. This is particularly pivotal for critical infrastructure such as the transportation networks, as damage or loss of functionality in these structures can lead to severe human and economic consequences, particularly following a disaster such as an earthquake. A widely adopted approach to analytically estimate the safety of a structure involves the comparison between seismic demand and structural capacity. The accuracy in the estimation of these two components can be highly improved using data collected from structural health monitoring systems. However, in most cases, the number of assets equipped with sensors is limited due to economic and logistical constraints. To address this challenge, a machine learning-based methodology is proposed for the reliability assessment of highway bridge networks. The methodology presented in this paper focuses on estimating the seismic capacity of reinforced concrete (RC) bridges and introduces a transfer learning framework that enables the extrapolation of information from sensor-equipped (monitored) bridges to those without instrumentation (non-monitored), utilizing a Bayesian network architecture. The accuracy of the proposed approach is evaluated by comparing the responses derived from finite element (FE) numerical simulations with those predicted by the machine learning model across the network.