67,000+ Australian Taxation Office documents as RAG-ready NDJSON/CSV — Edited Private Advice, public rulings and determinations, ATO Interpretative Decisions, practical compliance guidelines, taxpayer alerts, decision impact statements, practice statements and legislative instruments. Every document parsed into structured, typed fields for legal RAG, LLM fine-tuning, and tax research automation.

Machine-readable extracts of Australian Tax Office (ATO) legal documents, scraped from the ATO Legal Database. Each row is one ruling, decision, or guideline — fully parsed into structured columns (CSV) or JSON fields (NDJSON), with all multi-value entries pipe-delimited for downstream NLP, RAG ingestion, and legal AI over Australian tax law (ITAA 1997, ITAA 1936, GST Act).

Companion dataset: the legislation itself — the full text of the 22 ATO-administered Acts with per-section amendment history — is published separately at simplelex/Australian-Tax-Legislation-and-Amendment-History. The two are designed to be used together: the documents here cite provisions, and that dataset resolves them. They are separate repos because they come from different sources under different licences (ATO content is CC BY 3.0 AU; the Federal Register of Legislation is CC BY 4.0 with its own mandatory attribution wording).

Each dataset is published in two parallel files with identical content and field order:

JSON keys are identical to CSV column names — the schema tables below describe both formats.

• Dataset Splits
• Quick Start
• Schema
  • Edited Private Advice
  • ATO Interpretative Decisions
  • Public Rulings & Determinations
  • Practical Compliance Guidelines
  • Taxpayer Alerts
  • Decision Impact Statements
  • Law Administration Practice Statements
  • Legislative Instruments
  • Advice under Development
• Moved: the legislation datasets
• Use Cases
• Limitations
• Support the Project
• Source & Licensing
• Citation
• Terms of Service

Document counts are from the completeness audit of 29 July 2026, which verified every split against the live ATO Legal Database tree with zero gaps. They grow as the ATO publishes.

Looking for the legislation itself? The principal Acts and their amendment history moved to their own repo: simplelex/Australian-Tax-Legislation-and-Amendment-History. See Moved: the legislation datasets.

from datasets import load_dataset

# Use a named config — viewer-friendly, strongly typed schema
ds = load_dataset(
    "simplelex/ATO-Australian-Tax-Rulings-and-Guidance",
    "edited_private_advice",
    split="train",
)
print(ds[0])

# Available configs: edited_private_advice · ato_interpretative_decisions · public_rulings
#   practical_compliance_guidelines · taxpayer_alerts · decision_impact_statements
#   law_admin_practice_statements · legislative_instruments · advice_under_development

import pandas as pd

df = pd.read_json("edited_private_advice_all.jsonl", lines=True)
active = df[df["Is_Archived"] == False]
print(active[["Authorisation_Number", "Subject", "Date_of_Advice"]].head())

import pandas as pd

df = pd.read_csv("edited_private_advice_all.csv")
active = df[df["Is_Archived"] == False]
print(active[["Authorisation_Number", "Subject", "Date_of_Advice"]].head())

Multi-value fields (legislative references, ruling periods, related documents) use | as the internal delimiter in both formats:

# Explode pipe-delimited legislative references into one row per provision
provisions = (
    active["Relevant_Legislative_Provisions"]
    .str.split(" | ")
    .explode()
    .str.strip()
    .value_counts()
)
print(provisions.head(10))

The column tables below apply to both the CSV and NDJSON variants — every CSV column maps 1:1 to a JSON key of the same name in the corresponding .jsonl file, in the same order.

Private rulings issued to individual taxpayers, anonymised and edited for publication.

Edited summaries of Tax Office internal decisions, released under FOI obligations.

Binding public rulings (TR, TD, IT, CR, PR) and legislative instruments issued by the Commissioner.

Guidelines explaining how the ATO will administer specific areas of tax law, including safe harbours.

The ATO's early-warning notices about tax avoidance schemes currently under review. High practitioner value: warns clients away from flagged arrangements before audit action begins.

The ATO's official response to court and tribunal decisions — whether it accepts the outcome, is appealing, or will update its public guidance. Critical for practitioners: a favourable court ruling means nothing if the ATO is going to the High Court.

Internal ATO instructions to officers on how to apply the law in practice. Not legally binding on taxpayers but widely cited in audit correspondence and objection letters — practitioners use them to predict and contest ATO positions.

Legislative instruments, determinations, and notices made by the Commissioner under tax and related legislation. Registered on the Federal Register of Legislation (FRLI). Includes current instruments, repealed instruments, and drafts for comment.

The ATO's public register of the public advice and guidance it is currently developing — the pipeline behind the other splits. Sourced from the Advice under development program pages rather than the Legal Database. One row per advice product: a single [NNNN] issue often covers several products (draft Determination, final Ruling, Practice Statement) with separate expected completion dates, so Issue_Number is not a primary key — key on Source_URL.

Advice is removed from this split once the document it produces is published to the ATO Legal Database, so it always describes work still in progress.

The three legislation splits that used to live in this repo — frl_principal_legislation, frl_legislation_history and frl_legislation_with_history — are now published on their own at simplelex/Australian-Tax-Legislation-and-Amendment-History.

They were separated because they come from a different source under a different licence: ATO Legal Database content is © Commonwealth of Australia under CC BY 3.0 Australia, while the Federal Register of Legislation is CC BY 4.0 and carries its own mandatory attribution wording. A single repo can only declare one license:, so publishing them together mis-stated the legislation half.

The data, schema and column names are unchanged — only the repo and the config names:

• Fine-tuning tax LLMs — 67,000+ structured Australian tax law documents with rich reasoning fields (Reasons_for_Decision, ATO_View_of_Decision) make ready-made supervised fine-tuning examples for legal and tax domain LLMs; ideal training data for ITAA 1997 and GST Act specialists
• Legal RAG pipelines — each row is a self-contained retrieval unit; chunk on Ruling, Reasons_for_Decision, or Compliance_Approach for high-precision retrieval over Australian tax law; compatible with any vector store or RAG framework
• Tax research automation — query and filter across all ATO document types by Document_Type_Code, Is_Withdrawn, and Date_of_Issue to track policy evolution or surface current binding obligations automatically
• Legal NLP — classification, summarisation, and named-entity recognition on Australian tax law; well-suited to legal-NLP, domain-specific NLP research, and text-retrieval benchmarking
• Legislative graph construction — explode Legislative_References to build provision co-citation networks across ITAA 1997, ITAA 1936, GST Act, and more
• Compliance tooling — filter on Document_Type_Code + Is_Withdrawn == False to surface current binding obligations; combine with ATO_View_of_Decision in DIS records to identify contested positions
• Grounded answers over statute and guidance together — join Relevant_Legislative_Provisions here to Provision_Key in the legislation dataset so a generated answer can cite the ATO's interpretation and the provision text it interprets, with the amendment history behind it

• Currency — the ATO Legal Database is checked for new documents every weekday at 11 am AEST; datasets are updated automatically when new documents are detected. ATO documents may still be updated, withdrawn, or replaced between checks. Always verify against the ATO Legal Database for official or time-sensitive purposes.
• Coverage gaps — Edited Private Advice coverage begins in 2011; earlier private advice is not available on the ATO Legal Database. GST Industry Issues (GII) documents are excluded — they lack standard article structure and do not map cleanly to the schema.
• Parsing fidelity — content is extracted from HTML; some older documents (pre-2010) use non-standard markup that may produce blank or truncated fields. Complex tables, diagrams, and attached PDF appendices are not captured.
• Field sparsity — many optional fields (e.g. Date_of_Withdrawal, Date_Scheme_Commenced) are blank for the majority of records, reflecting source documents that simply do not include those details.
• Boolean encoding — Is_Archived, Is_Withdrawn, Is_Draft, and similar flags are stored as the strings "True" / "False" in both CSV and NDJSON for round-trip parity; cast explicitly if your pipeline expects native booleans.
• Australian jurisdiction only — all documents are issued by the Australian Taxation Office and relate exclusively to Australian tax law (ITAA 1997, ITAA 1936, GST Act, and related legislation). This dataset is not suitable as a general-purpose legal corpus for other jurisdictions.
• Not legal advice — this dataset reproduces publicly available ATO materials for informational and research purposes only. It is not a substitute for professional legal or tax advice.

This dataset is free. If it saves you time on data engineering, a coffee is appreciated:

☕ Buy Me a Coffee

Source: the ATO Legal Database (ato.gov.au), © Commonwealth of Australia, made available under the Creative Commons Attribution 3.0 Australia licence.

The licence permits commercial use, redistribution and adaptation with attribution. This structured compilation is published by SimpleLex under the same Creative Commons terms.

Legislation sourced from the Federal Register of Legislation is published in the companion repo under CC BY 4.0, with the FRL's own required attribution wording — it is licensed differently from the ATO content here, which is why the two are separate datasets.

@dataset{simplelex_ato_rulings_2026,
  author    = {SimpleLex},
  title     = {ATO Rulings \& Guidance — Australian Tax Law, Structured for AI},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/simplelex/ATO-Australian-Tax-Rulings-and-Guidance}
}

Last updated: July 2026

A structured dataset of publicly available legal documents published by the Australian Taxation Office (ATO) at ato.gov.au. Documents include Edited Private Advice, Public Rulings, Determinations, ATO Interpretive Decisions, Practical Compliance Guidelines, Taxpayer Alerts, Decision Impact Statements, Law Administration Practice Statements, legislative instruments, and the ATO's register of advice still under development. All datasets are compiled into machine-readable CSV and NDJSON format.

The principal Acts the ATO administers, with reconstructed amendment-history notes sourced from the Federal Register of Legislation, are published as a separate dataset under its own licence and terms.

You may use this dataset for any lawful purpose, including research, analysis, building applications, and commercial projects. You do not need to ask permission. Attribution is appreciated but not required for the compilation itself.

The underlying ATO documents are published under the Creative Commons Attribution 3.0 Australia licence. The Commonwealth retains copyright in the original materials. This dataset is a structured compilation and does not claim copyright in the original source material.

Nothing in this dataset constitutes legal or tax advice. Documents are reproductions of publicly available ATO materials provided for informational purposes only. Always consult a registered tax agent or legal professional for advice about your specific situation.

We make no warranties about the completeness, accuracy, or currency of the data. ATO documents may be updated, withdrawn, or replaced after the dataset was last scraped. Always verify against the ATO Legal Database for official or time-sensitive purposes.

To the maximum extent permitted by Australian law, simplelex accepts no liability for any direct, indirect, or consequential loss arising from use of or reliance on this dataset.

We may update these terms at any time. Continued use of the dataset following an update constitutes acceptance of the revised terms.