add federal law: the whole United States Code, 60,433 sections across all 53 titles, measured (OLRC USLM XML bulk, release point Public Law 119-103, 2026-09-02). Notes/annotations excluded from text; source credit, positive-law flag and repealed/omitted status kept in extra. Card regenerated with a federal row and a per-title table.

README.md

@@ -3,8 +3,8 @@ license: cc0-1.0
 language: [en]
 task_categories: [text-retrieval, question-answering]
 size_categories: [1M<n<10M]
-tags: [legal, law, statutes, state-law, us-law, public-domain, rag, legal-nlp]
-pretty_name: US State Statutes (verbatim)
+tags: [legal, law, statutes, us-code, federal-law, state-law, us-law, public-domain, rag, legal-nlp]
+pretty_name: US Statutes — federal and state (verbatim)
 configs:
 - config_name: all
   data_files: data/*/statutes.jsonl.gz
@@ -38,14 +38,18 @@ dataset_info:
     dtype: string
 ---
 
-# US State Statutes — held word for word
+# US Statutes — held word for word
 
-**1,073,897 statute sections** across **24 states**, each section as its legislature publishes it,
-with the URL it was taken from. Statutes are the law a court applies; a paraphrase is not usable in a
-filing, so the text here is verbatim.
+**1,134,330 statute sections**: the whole **United States Code** (60,433 sections, all 53 titles)
+plus **24 states** (1,073,897 sections), each section as its legislature publishes it, with
+the URL it was taken from. Statutes are the law a court applies; a paraphrase is not usable in a filing,
+so the text here is verbatim.
 
-| state | sections | codes |
+Ask *"what does 42 U.S.C. § 1983 say"* and this dataset answers with the section, not a summary of it.
+
+| jurisdiction | sections | codes |
 |---|--:|---|
+| **United States (federal)** (`us`) | **60,433** | `United States Code`, all 53 titles |
 | California (`ca`) | 162,431 | `Government Code`, `Health and Safety Code`, `Education Code` … |
 | Texas (`tx`) | 122,681 | `Special District Local Laws Code`, `Government Code`, `Health & Safety Code` … |
 | Oregon (`or`) | 72,649 | `Oregon Revised Statutes` |
@@ -76,38 +80,146 @@ taken from any index page's own total.
 
 `*` Illinois is a **snapshot of a harvest that was still running** when its file was published —
 48,587 sections so far (the ILCS crawl at ftp.ilga.gov honours a 10-second crawl delay). It is a
-complete record of what had been fetched, and will be republished when the crawl finishes. Nebraska was
-published that way earlier and is now the **complete** code, all 55,413 sections. New York is the
+complete record of what had been fetched, and will be republished when the crawl finishes. New York is the
 Consolidated Laws only — Unconsolidated Laws, Court Acts and the miscellaneous categories at nysenate.gov
-were out of scope for this pass.
+were out of scope for that pass.
+
+## The United States Code (`jurisdiction: us`, `level: federal`)
+
+Taken from the **Office of the Law Revision Counsel** of the U.S. House of Representatives, the statutory
+publisher of the Code, at its **release point Public Law 119-103 (2026-09-02)** — one bulk download of the
+official USLM XML (`xml_uscAll@119-103.zip`), not a crawl. The Code runs to Title 54, but there is no
+Title 53 — the OLRC prints it as `Title 53 [Reserved]` — so these 53 titles are the whole Code.
+
+| title | name | sections | positive law |
+|--:|---|--:|---|
+| 1 | General Provisions | 39 | yes |
+| 2 | The Congress | 1,612 | no |
+| 3 | The President | 64 | yes |
+| 4 | Flag and Seal, Seat of Government, and the States | 47 | yes |
+| 5 | Government Organization and Employees | 1,163 | yes |
+| 6 | Domestic Security | 569 | no |
+| 7 | Agriculture | 2,926 | no |
+| 8 | Aliens and Nationality | 376 | no |
+| 9 | Arbitration | 33 | yes |
+| 10 | Armed Forces | 4,033 | yes |
+| 11 | Bankruptcy | 280 | yes |
+| 12 | Banks and Banking | 1,999 | no |
+| 13 | Census | 70 | yes |
+| 14 | Coast Guard | 449 | yes |
+| 15 | Commerce and Trade | 2,701 | no |
+| 16 | Conservation | 5,095 | no |
+| 17 | Copyrights | 153 | yes |
+| 18 | Crimes and Criminal Procedure | 1,399 | yes |
+| 19 | Customs Duties | 1,111 | no |
+| 20 | Education | 2,029 | no |
+| 21 | Food and Drugs | 757 | no |
+| 22 | Foreign Relations and Intercourse | 3,475 | no |
+| 23 | Highways | 172 | yes |
+| 24 | Hospitals and Asylums | 162 | no |
+| 25 | Indians | 2,399 | no |
+| 26 | Internal Revenue Code | 2,161 | no |
+| 27 | Intoxicating Liquors | 45 | no |
+| 28 | Judiciary and Judicial Procedure | 806 | yes |
+| 29 | Labor | 844 | no |
+| 30 | Mineral Lands and Mining | 626 | no |
+| 31 | Money and Finance | 498 | yes |
+| 32 | National Guard | 78 | yes |
+| 33 | Navigation and Navigable Waters | 1,569 | no |
+| 34 | Crime Control and Law Enforcement | 795 | no |
+| 35 | Patents | 175 | yes |
+| 36 | Patriotic and National Observances, Ceremonies, and Organizations | 1,173 | yes |
+| 37 | Pay and Allowances of the Uniformed Services | 234 | yes |
+| 38 | Veterans' Benefits | 1,198 | yes |
+| 39 | Postal Service | 181 | yes |
+| 40 | Public Buildings, Property, and Works | 459 | yes |
+| 41 | Public Contracts | 241 | yes |
+| 42 | The Public Health and Welfare | 8,481 | no |
+| 43 | Public Lands | 1,464 | no |
+| 44 | Public Printing and Documents | 356 | yes |
+| 45 | Railroads | 294 | no |
+| 46 | Shipping | 1,032 | yes |
+| 47 | Telecommunications | 432 | no |
+| 48 | Territories and Insular Possessions | 661 | no |
+| 49 | Transportation | 1,371 | yes |
+| 50 | War and National Defense | 1,360 | no |
+| 51 | National and Commercial Space Programs | 249 | yes |
+| 52 | Voting and Elections | 171 | no |
+| 54 | National Park Service and Related Programs | 366 | yes |
+
+**27 titles are positive law** and **26 are not**; the distinction is carried per row in
+`extra.is_positive_law`, read from the XML's own `<property role="is-positive-law">` and checked against the
+star the OLRC prints beside positive-law titles on its download page — the two agree on all 27. In a non-positive-law
+title the Code text is *prima facie* evidence of the law and the underlying Statutes at Large are legal
+evidence, so the distinction matters when you cite.
+
+What is in `text`, and what is deliberately not:
+
+- **In:** the section's own number and heading, and its operative text — subsections, paragraphs, clauses,
+  provisos, in order.
+- **Out:** everything USLM marks as an annotation. Editorial and statutory notes, amendment notes,
+  codification notes, effective-date notes and the three-column tables are **not statute text** and are not
+  mixed into `text`. The statutory **source credit** — `(R.S. § 1979; Pub. L. 96–170 …)` — is kept, but in
+  `extra.source_credit`, not in the text of the law. Footnote reference marks are dropped from `text` and
+  the footnotes themselves are kept in `extra.footnotes`.
+- **Retained and marked:** sections that are repealed, omitted, transferred, renumbered, reserved or vacant
+  are kept — a repealed section is often exactly what a lawyer is looking for — and flagged in
+  `extra.status` (repealed 5,131, omitted 2,037, transferred 1,911, renumbered 368, vacant 17, reserved 16, unknown 1).
+  Sections with no such flag are current (50,952).
+- **Not here:** the US Code *appendices* (Titles 5, 11, 18, 28 and 50 Appendix). Those carry the Federal
+  Rules of Civil Procedure, Appellate Procedure, Evidence, Criminal Procedure and Bankruptcy Procedure,
+  which are court rules rather than statutes and live in `docketx/court-rules`.
+
+Six section numbers are used twice in their own title — a real anomaly in the Code, not a parsing error.
+Congress has enacted two § 1932s of title 28 (Judicial Panel on Multidistrict Litigation, and Revocation
+of earned release credit), and the same has happened at 5 U.S.C. § 3598, 5 U.S.C. § 5757,
+10 U.S.C. § 130g, 38 U.S.C. § 1167 and 40 U.S.C. § 3318. Both sections of each pair are published, with an
+identical `citation` and `id`s distinguished by a `_d` suffix; the Code itself footnotes the clash and that
+footnote is in `extra.footnotes`.
 
 ## The record
 
 Every row has the same twelve columns, all strings:
 
-`id`, `level` (`state`), `jurisdiction`, `code`, `section`, `citation`, `title`, `text` (verbatim),
-`source` (the URL), `license`, `retrieved_at`, `extra`.
+`id`, `level` (`federal` or `state`), `jurisdiction` (`us`, or the two-letter state), `code`, `section`,
+`citation`, `title`, `text` (verbatim), `source` (the URL), `license`, `retrieved_at`, `extra`.
 
 `extra` is a JSON **string** (parse it with `json.loads`) carrying whatever else the source exposed —
-title, chapter, subchapter, catchline, history note, edition — so that it can differ by state without
-breaking the shared schema.
+title, chapter, subchapter, part, catchline, history note, edition — so that it can differ by jurisdiction
+without breaking the shared schema. Federal rows add `title_number`, `title_name`, `is_positive_law`,
+`identifier` (the USLM path, e.g. `/us/usc/t42/s1983`), `release_point`, `source_credit`, `status` where
+one applies, and `browse_url` (the OLRC viewer page for that section).
 
 ## Load it
 
 ```python
 from datasets import load_dataset
-ds = load_dataset("docketx/us-statutes")                     # all states
-al = ds["train"].filter(lambda r: r["jurisdiction"] == "al") # one state
+ds = load_dataset("docketx/us-statutes")                       # federal + every state
+usc = ds["train"].filter(lambda r: r["level"] == "federal")    # the United States Code
+al  = ds["train"].filter(lambda r: r["jurisdiction"] == "al")  # one state
 ```
 
 ## Coverage, stated honestly
 
-These are the states whose statutes we have verified against a live official source, with the structure
-proven by extracting a real section before any harvest ran. A state missing from this table is one we
-have **not yet verified** — in several cases because the state licensed its code to a commercial
+Federal statutory law is complete: every title of the US Code at the release point named above.
+
+The states listed are the ones whose statutes we have verified against a live official source, with the
+structure proven by extracting a real section before any harvest ran. A state missing from this table is
+one we have **not yet verified** — in several cases because the state licensed its code to a commercial
 publisher and no official machine-readable source exists. It is never a state whose law we hold and
 withheld. Coverage grows as each state clears that check.
 
-Public-domain government works. Part of the [DocketX](https://docketx.ai) open legal corpus, alongside
-`docketx/us-caselaw` (10.8M opinions), `docketx/court-rules`, `docketx/oral-arguments-us` and
-`docketx/us-judges`.
+## Source and credit
+
+The United States Code is prepared and published by the **Office of the Law Revision Counsel of the U.S.
+House of Representatives**, which is the publisher Congress designated for it. It is a work of the United
+States Government and is **in the public domain** — not copyrighted, and free of any licence condition. The
+same data is mirrored by the **U.S. Government Publishing Office** on govinfo.gov. Nothing here is
+republished from a commercial annotator: there are no headnotes, no editorial annotations and no
+publisher's notes in this dataset.
+
+State codes are published by the named state legislature, legislative service or code commission, likewise
+as government works. Each row carries the exact URL it came from in `source`.
+
+Part of the [DocketX](https://docketx.ai) open legal corpus, alongside `docketx/us-caselaw`
+(10.8M opinions), `docketx/court-rules`, `docketx/oral-arguments-us` and `docketx/us-judges`.

data/us/statutes.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da85535e02afa3bfcd28f4977377258f55b62642adc8e83602659aaa89c40931
+size 43362888