Provenance Laundering in Longitudinal LLM Workflows
Provenance Laundering in Longitudinal LLM Workflows introduces a process-level failure in which information gains practical authority through repetition, inheritance, compression, or retrieval without a corresponding increase in external evidence. The characteristic sequence is: generated → repeated → familiar → treated as prior state → treated as established. The core failure is an authority lift without evidence lift. The paper develops provenance laundering independently from hallucination, false execution status, source-monitoring error, the illusory truth effect, memory poisoning, regulatory mimicry, and KORG. It proposes operational stages, an evidence-authority delta, a detection test, recovery protocol, falsification conditions, and a process distinction between lineage degradation and higher-order flow capture. The construct emerged from longitudinal naturalistic human-LLM work within the AI Working Method research program. The paper describes observable process relations and does not claim an internal model or code-level mechanism.
Authors
- Alen Širola
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-13
- DOI
- https://doi.org/10.5281/zenodo.24204594
- Citations
- 1
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- article
- Field-Weighted Citation Impact
- 1.39