CAPTCHAs play critical roles in maintaining the security of various Web services by distinguishing humans from automated programs and preventing Web services from being abused. CAPTCHAs are designed to block automated programs by presenting questions that are easy for humans but difficult for computers, e.g., recognition of visual digits or audio utterances. Recent audio CAPTCHAs, such as Google's audio reCAPTCHA, have presented overlapping and distorted target voices with stationary background noise. We investigate the security of overlapping audio CAPTCHAs by developing an audio reCAPTCHA solver. Our solver is constructed based on speech recognition techniques using hidden Markov models (HMMs). It is implemented by using an off-the-shelf library HMM Toolkit. Our experiments revealed vulnerabilities in the current version of audio reCAPTCHA with the solver cracking 52% of the questions. We further explain that background stationary noise did not contribute to enhance security against our solver.