Abstract The assumption that deleted digital information is irretrievably lost remains widespread, yet this belief obscures the continued existence of residual traces within archives, backups, and systems that implement only partial or ‘soft’ deletion. Such remnants, though frequently overlooked, may persist in shaping the operation of artificial intelligence (AI) and therefore warrant critical examination. This article addresses three interrelated questions. First, through what mechanisms do obsolete or concealed data fragments re-enter AI systems and exert influence that manifests as bias or distortion? Second, in what ways might the trajectory of digital information be conceptualised through reference to ecological processes such as decomposition, renewal, and systemic adaptation? Third, what forms of regulatory or procedural innovation – here articulated as a model of ‘digital composting’ – could facilitate the identification, evaluation, and responsible management of residual data? The analysis demonstrates that outdated data frequently function as ‘ghost inputs’. Despite their invisibility, these elements shape the generative capacity of AI, modify the narratives it produces, and subtly recalibrate public discourse as well as shared cultural memory. Their persistence underscores the communicative and social significance of digital traces once presumed to be obsolete. To advance this discussion, the article introduces the concept of a ‘Data Decay Pathway’. This framework offers a novel means of addressing accountability and transparency in digital systems, emphasising that processes of digital decay, much like those observed in natural ecosystems, may serve either to sustain the vitality of collective memory or to propagate forms of distortion and systemic vulnerability.
Hanna Gaweł (2026) studied this question.