Replication data for "Rhetoric Versus Reality: An NLP Analysis of AI-Adoption Language and Effectiveness Ratings in GAFILAT Mutual Evaluation Reports"

# Replication Package — Rhetoric Versus Reality: An NLP Analysis of AI-Adoption Language and Effectiveness Ratings in GAFILAT Mutual Evaluation Reports Alfredo Agustín Merlet EchavarríaUniversidad de Sevilla | ORCID: 0009-0007-8558-164X ## Contents - `raw_text_corpus/` — full plain-text extraction of all 18 GAFILAT Fourth Round Mutual Evaluation Reports (MERs), one file per country, named `{Country}_MER{Year}.txt`. Extracted via `pdftotext -layout`, no truncation, from official PDF sources.- `Search_Protocol.txt` — full documentation of the lexical search protocol: exact regex patterns used, per-language term families (Spanish for 16 documents, English for Chile and Uruguay), and the manual-verification procedure applied to every match.- `Table_I_results_FINAL.csv` — country-level results table: MER year, page count, and counts of AI-language, "algoritmo"/"algorithm", and "automatiz*"/"automat*" mentions, with notes on borderline classification decisions. ## Summary of findings Of the 18 GAFILAT member states' Mutual Evaluation Reports, exactly 2 (Brazil and Chile)contain explicit descriptions of functional AI/machine-learning systems within theirfinancial intelligence units. The remaining 16 contain no AI-related terminology, despitesubstantial general-automation vocabulary in several (Uruguay: 21 mentions; El Salvador: 11;Peru: 11). Only two additional countries (Paraguay, Nicaragua) use the word"algoritmo"/"algorithm" outside these two cases, in both instances describing non-analyticaluses (rule-based scoring and encryption, respectively). ## Citation If you use this dataset, please cite: Merlet Echavarría, A.A. (2026). Replication data for "Rhetoric Versus Reality: An NLPAnalysis of AI-Adoption Language and Effectiveness Ratings in GAFILAT Mutual EvaluationReports" [Data set]. Zenodo. https://doi.org/[DOI TO BE ASSIGNED ON DEPOSIT] ## License Raw text extractions are derived from public documents published by GAFILAT(https://biblioteca.gafilat.org) and national financial authorities of Chile and Panama.Original documents © GAFILAT and respective national authorities. This derived plain-textcorpus and accompanying analysis files are shared for research replication purposes. ## Source documents Full list of original PDF sources (with URLs) is provided in `Search_Protocol.txt`, Section 1.