We propose an interactive hidden Markov chain model for assessing credit risk. Two types of states exist in a hidden Markov chain model: hidden states and observable states. Hidden states can be used to incorporate the risk associated with the economic state. Our model incorporates the interactive effect of observable states and hidden states, which is more realistic and not included in traditional hidden Markov models. We also transform the proposed model into a risk-neutral interactive hidden Markov model using a time-varying risk premium. Finally, the 3 simulated results indicate that for investment grade probability, facing a normal risk state next year is higher than that of speculative grades. However, medium-risk grade probability cannot be determined because of the uncertainty of the 3 simulations. Thus, banking must focus on borrowers’ rating classes when making a lending decision.