Recent studies have shown that stimulus history can be decoded via the use of broadband sensory impulses to reactivate mnemonic representations. It has also been shown that predictive mechanisms in the auditory system demonstrate similar tonotopic organization of neural activity as that elicited by the perceived stimuli. However, it remains unclear if the mnemonic and predictive information can be decoded from cortical activity simultaneously and from overlapping neural populations. Here, we passively exposed anesthetized rats to complex auditory sequences consisting of artificial vowel triplets, while recording electrocorticography (ECoG) data from the auditory cortex. Occasionally, vowels were replaced by noise bursts, i.e., a completely uninformative stimulus. Although the noise bursts did not carry any information about the memory or prediction of stimuli, we could decode both mnemonic and predictive information from neural activity evoked by the bursts, showing that sensory cortical networks maintain both mnemonic and predictive representations independently of the currently processed sensations. Crucially, we also demonstrate that predictive representations are learned over the course of stimulation at two distinct time scales, reflected in two dissociable time windows of neural activity. These results are novel in that they show, for the first time, that during exposure to dynamically changing sequences, information about memory and prediction can be decoded from neural activity in the auditory cortex. More critically, they also show that the predictive representations are rapidly acquired and dynamically updated to match the stimulus statistics. Strikingly, these effects are observed under full anesthesia, indicating that learning of complex and rapidly changing contingencies can occur under passive stimulation and without awareness, providing novel evidence for the automaticity of memory and predictive computations. This largely empirical study will be of interest for computational neuroscientists working in the fields of predictive coding, population coding, memory encoding, and statistical learning.