Many reinforced concrete bridges in Europe and around the world are damaged by reinforcement corrosion and the annual maintenance costs are enormous. It is therefore important to develop reliable methods to assess the structural capacity of corroded reinforced concrete structures and avoid unnecessary maintenance costs. Although there are advanced models for determining the load carrying capacity of structures, it is not obvious how they should be used to verify the performance of existing structures. To confidently assess the bond of corroded reinforcement in concrete, for example, the calculation model must give a sufficient safety margin. When designing new structures, semi-probabilistic approaches (such as the partial safety factor method) are adopted to achieve the target reliabilities specified in structural design codes. This paper uses probabilistic methods to develop partial factors for application in an existing bond model, to assess the safety of corroded reinforced concrete structures. The response of the bond model was studied using Monte Carlo (MC) simulations for several design cases, with probability distributions fitted to the results. Partial factors were then derived, based on these distributions. Furthermore, an MC-based simulation technique called “importance sampling” was used to study the reliability of several deterministic bond assessments conducted using these partial factors. The results show that deterministic assessments which use the proposed partial factors lead to a safety level at least equal to the target value. The results presented in this paper will support the assessment of reinforced concrete structures with anchorage problems and give a reasonable approximation of the anchorage capacity with sufficient safety margin. When generalised to cover other failure modes and structural configurations, this will enable better utilisation of damaged structures and lead to major environmental and economical savings for society.