# Licence

Issued 2026-08-24, v5. Earlier versions of the pages said "all rights reserved" while
telling readers to download and re-run everything, which does not work as law. Fixed:

- **Code** (every `.py` file in this bundle): MIT Licence.
- **Data and text** (the CSVs, the registers, the standards, the pages): Creative Commons
  Attribution 4.0 (CC BY 4.0). Reuse freely with attribution to patrickneilbradley.com.
  **Carve-out (2026-09-02):** `style_corpus_all.csv` reproduces short samples of
  third-party text (candidates' social posts, quotations and releases) for the purpose of
  the writing-measurement analysis, each with its source URL. That text is the property of
  its authors or is in the public domain (official house.gov releases); it is NOT relicensed
  here and CC BY 4.0 does not extend to it. The grant covers this project's compilation,
  coding and annotations of that corpus, not the quoted text itself.
- **FEC source data** is a work of the United States government, public domain; this
  bundle's finance figures are derived from it.
- The candidate information in `roster_115.csv` and `phase2.py` describes public figures
  from cited public sources under a written evidentiary standard; the licence covers the
  compilation, not the underlying facts, which belong to no one.

MIT Licence text: https://opensource.org/license/mit
CC BY 4.0 text: https://creativecommons.org/licenses/by/4.0/

---

## What CC BY 4.0 does not cover, and what this dataset asks of you anyway

Added 2026-08-30 after a review session briefed as a media lawyer (an AI session run by the author, as every review of this project was — see the predictions page) pointed out that a reuser looks here
and nowhere else, and that three things a reuser of *this* dataset needs were missing.

**1. This is personal data about 115 named living people.** CC BY 4.0 §2(b) expressly does
*not* license publicity, privacy or personality rights, and it does not license moral
rights. Downloading this file makes you an independent controller under UK/EU data
protection law, with your own obligations — including Article 9 (the `faith` column is special-category data, filled only from the subjects' own public statements or documented membership, and blank otherwise) and Article 10, which restricts
processing of criminal-conviction data. The licence gives you copyright permission. It
gives you nothing else, and it cannot.

**2. No endorsement.** Attribution is required; implying that this project endorses your
use of the data is not permitted, and under CC BY 4.0 §3(a)(1)(A)(iii) attribution must be
removed on request.

**3. The disposition travels with the charge. This is a condition of honour and it is the
most important line in this file.**

> Every legal matter in this dataset is published with its outcome. The `legal` column
> alone is not a record — it is an insinuation. **Do not reuse `legal` without
> `legal_disposition` and `legal_detail` in the same view.** A pardon, a nolle prosequi, a
> dismissal for want of prosecution and a reduced charge all read as "something happened"
> once the outcome column is dropped by a join, and three of the people in this file would
> be misrepresented by that in a way no correction reaches.
>
> Any row carrying `tier_verified = no` (as shipped there are none — every tier claim in
> this release is supported by its citation) is **not established to this index's own
> standard** and must not be presented as established. Likewise any row whose `legal` value
> reads *"Not established — see detail"* is not a finding; the index declines to assert it and
> so must you. These flags are the reader-protection: honour them if a future release sets them.

CC BY 4.0 cannot compel any of this. It is asked in the strongest terms available, and the
dataset is structured — outcome as its own column, not as free text — so that honouring it
costs a reuser nothing but reading this paragraph.