pii-guard
Copyright 2026 the pii-guard contributors

This product is licensed under the Apache License, Version 2.0 (see LICENSE). It includes
data from the third parties below, each under its own licence. The source, licence and
SHA-256 of every committed lexicon file are in data/lexicons/MANIFEST.md (ADR 0009).

------------------------------------------------------------------------------------------
Shipped in the Python package (src/pii_guard/lang/data/hu/)
------------------------------------------------------------------------------------------

Hungarian Central Statistical Office (Központi Statisztikai Hivatal, KSH)
  Licence: Creative Commons Attribution 4.0 International (CC BY 4.0),
           https://creativecommons.org/licenses/by/4.0/ ; terms: https://www.ksh.hu/copyright
  - settlements.tsv, postcodes.tsv: settlement names, counties, KSH codes, populations and
    postcode-to-settlement pairs, from the KSH Helységnévtár 2026-01-01
    (https://www.ksh.hu/docs/helysegnevtar/hnt_letoltes_2026.xlsx).
  - recent_given_names_f.tsv, recent_given_names_m.tsv: the most chosen given names of
    newborns, from the KSH "Gyermekek utónevei" infographics, 2000-2025.
  Changes: extracted, normalised and reduced to the columns listed; see
  data/lexicons/fetch.py.

OpenStreetMap
  © OpenStreetMap contributors. Licence: Open Database License 1.0 (ODbL 1.0),
  https://opendatacommons.org/licenses/odbl/1-0/ ; data via Geofabrik
  (https://download.geofabrik.de/europe/hungary-latest.osm.pbf).
  - street_names.tsv, street_types.tsv: the 100 most common Hungarian street names and the
    counts of public-space types, a non-substantial extract of highway=* name values.

------------------------------------------------------------------------------------------
Optional dependency, installed in the Docker images (not part of the pii-guard package)
------------------------------------------------------------------------------------------

meta-sam-parser 0.1.0 (Meta's official SAM 3.1 output parser)
  Copyright (c) Meta Platforms, Inc. and affiliates. https://github.com/meta-models/meta-sam
  Licence: SAM License (not an open-source licence); full text in
  licenses/meta-sam-parser-SAM-LICENSE.txt.
  - Installed by the optional `sam` extra and included in the Docker images (Dockerfile,
    Dockerfile.hosted). pii-guard uses it unmodified; without it, SAM 3.1 output is read by
    pii-guard's own decoder.

------------------------------------------------------------------------------------------
In the repository only (evaluation fixtures, not in the Python package or the image)
------------------------------------------------------------------------------------------

Ai4Privacy OpenPII 1M
  Copyright © 2026 Ai Suisse SA. Attribution: Ai4Privacy / Ai Suisse SA,
  https://huggingface.co/datasets/ai4privacy/pii-masking-openpii-1m
  Licence: Creative Commons Attribution 4.0 International (CC-BY-4.0).
  - eval/fixtures/public/*-ai4privacy.jsonl: a seeded sample of the validation split,
    relabelled to pii-guard's entity types (eval/fixtures/public/README.md).

AI-generated fictional specimen images
  - eval/fixtures/visual/images/: portraits, signatures, stamps and textures generated with
    Muse Image (muse-image-1.0, Meta Model API); the people shown do not exist. Layouts and
    every value on them are fictional. Details: eval/fixtures/visual/NOTICE.

Evaluation baselines (Presidio, LLM Guard, PrivAiTe, Eka PII Redaction and their models)
are downloaded at run time into their own environments and are not redistributed. Their
licences, some of them non-commercial or AGPL-3.0, are listed in docs/evaluation.md.
