An open, versioned, schema-validated registry of data brokers and how to opt out of each one — the facts, the editorial judgment (how automatable is this opt-out? how urgent?), and the build pipeline that compiles them into a single artifact. This is a dataset repo, not an app: it has no UI of its own on purpose, so it can be consumed by more than one frontend, run locally, or forked independently.
The reference frontend that consumes this dataset is
Make Yourself Expensive, built by
Transparency Cascade Press — a free, no-account,
no-server tool that walks a reader through opting out of every broker in here by hand, the
way the Consumer Reports study found beats every paid service.
Nothing about this dataset depends on that frontend, though; anyone is welcome to build a
different one against the same data/brokers.json artifact.
Each broker is authored as a markdown file with YAML frontmatter under data/brokers/, validated
against broker.schema.json, and compiled into data/brokers.json by scripts/build_brokers.py.
Every entry carries an authored tier (auto/assisted/guided — how much the tool can automate)
and priority (crucial/high/standard — impact), plus a last_verified date: null means a
human hasn't yet confirmed the routing against the broker's own page, and the app badges those
entries as unverified.
Sources this dataset draws on and credits:
- Data Brokers Watch — initial roster baseline.
- Privacy Rights Clearinghouse — data broker registry.
- Yael Grauer's Big Ass Data Broker Opt-Out List (BADBOOL).
- Transparency Cascade Press's own verification research, tiering, and instructions.
- California's free state DROP tool (linked, not bundled).
Brokers not yet given a full editorial pass are tracked in data/backlog.json rather than silently
dropped or guessed at — _meta.authored_count / backlog_count / total_known in the compiled
dataset show real coverage.
scripts/build.sh
runs the full pipeline in order: scripts/check_links.py (refreshes data/link-status.json
liveness cache by reading URLs straight from data/brokers/*.md) then scripts/build_brokers.py
(validates every broker file against broker.schema.json, merges in liveness, computes a content
hash, and writes data/brokers.json + data/manifest.json).
A daily GitHub Action (.github/workflows/page-monitor.yml) runs scripts/monitor_pages.py,
which fetches every opt-out page, keeps a text snapshot of each under monitor/snapshots/, and
files one page-monitor issue per broker whose page changed (text, form fields, CAPTCHA,
redirects) or disagrees with its entry. Those issues are the re-verification queue: a changed page
means "look again", never "the entry is wrong". It never edits data/.
To validate a single file while authoring (e.g. as a pre-commit check):
python3 scripts/validate_broker.py data/brokers/some-broker.md
This repo has no UI of its own. Make Yourself Expensive
is the reference frontend (on-device profile, tiered opt-out actions, dataset-integrity
checking against the content_hash this repo's build emits) — but any tool can consume the
compiled data/brokers.json artifact the same way it does. A second or third frontend is
exactly as welcome as the first.
See CONTRIBUTING.md — most contributions are adding or correcting a broker entry, which
doesn't require touching any code.
This repository uses two licenses, on purpose — a dataset and the software that reads it are different things:
- App code (scripts, build tooling): MIT. See
LICENSE. - Broker dataset (
data/brokers/, compileddata/brokers.json,broker.schema.json): CC BY-NC-SA 4.0, crediting Privacy Rights Clearinghouse, Yael Grauer/BADBOOL, and Data Brokers Watch. SeeLICENSE-DATAfor the full attribution and scope notes.