Replication Package for From Risk Registers to Uncertainty Observatories: Managing Shadow Generative AI Innovation Under High Risk and High Uncertainty

This dataset supports the manuscript From Risk Registers to Uncertainty Observatories: Managing Shadow Generative AI Innovation Under High Risk and High Uncertainty. It documents the empirical basis for a three-part analysis of shadow generative AI governance: (1) literature-metadata synthesis, topic modeling, and keyword co-occurrence; (2) public adoption and task-exposure indicators from BTOS, Stack Overflow, and the Anthropic Economic Index; and (3) rule-coded AI provider-status incidents used as weak-signal evidence. The deposit includes processed datasets, analysis scripts, generated tables and figures, the data manifest, codebook, methodological notes, search-query documentation, checksums, and excluded-data notes. The authors assembled the source inventory, screened and deduplicated the literature metadata, cleaned and harmonized public datasets, developed the variable mappings, coded incident categories, ran the statistical models and robustness checks, and curated the manuscript-facing outputs. These materials are intended to support transparent review of the article's public proxy-evidence claims rather than to estimate hidden workplace AI use directly. Raw third-party files are not redistributed where copyright, platform terms, or licensing conditions are uncertain. The manifest and excluded data notes identify the original public source locations and explain which processed or aggregate artifacts are provided in their place.