The acquisition of new orthographic representations is a rapid and highly automatic process in monolingual readers. Our study extends existing research to biliterate populations, addressing the impact of phonological inconsistencies across native (L1) and second language (L2) alphabets during orthographic learning. Behavioral and EEG signals were collected from a group of 24 Russian-English biliterates via a reading-aloud task using familiar and novel words repeated across ten consecutive blocks in three Script conditions: (1) native Cyrillic, (2) non-native Roman, and (3) ambiguous (phonologically inconsistent graphemes shared by L1 and L2 alphabets). Linear mixed-effects modelling of both behavioral and ERP data revealed reliable Block x Lexicality x Script interactions, indicating that naming latencies and brain activity changed differently across training blocks for novel and familiar words and, importantly, depending on script presentation. Particularly, novel words presented in the ambiguous script showed longer naming latencies and slower reading automatization than those presented in L1 and L2 alphabets. Nonetheless, despite this interference, their naming latencies matched those of familiar words before the end of the training, suggesting the attribution of their representations in the reader's lexicon. The enhancement of early brain responses observed for these stimuli alongside their training confirmed the improvement in their orthographic analysis and lexical access. Critically, this pattern of results was not found for familiar, already represented words, which exhibited a suppression of their brain activity across repetitions. Overall, our results indicate that phonological inconsistency interferes with novel word encoding but it does not prevent efficient attribution of orthographic representations.