Community Bans — Terms of use

v1.1 · effective 2026-09-30 · what changed

These terms apply to the unmask.sh Community Bans service: the endpoints at https://unmask.sh/api/feed/* and the published file https://unmask.sh/api/feed/list.json. By submitting reports, votes, or comments (= using the Submit opt-in or the inline actions on the Bans page), you accept these terms. Subscribing (= subscribe_mode set to fetch or fetch_apply, pulling read-only data) does not require acceptance. unmask subscribes by default; see §6 and the privacy policy.

1. What you may submit, vote on, or comment on

IP address, JA4 fingerprint, free-text reason, and a free-text comment (≤ 280 chars) tied to a BAN button press in your own unmask installation.
👍 / 👎 votes on other installs' submissions and follow-up comments (≤ 280 chars) when you have first-hand information about the reported traffic.
A 2-letter ISO country code for your install's location, which the hub looks up from your install's IP address while the publish_country setting is on (default ON; see the privacy policy for what turning it off does). Never the reported client's country.
Submissions / votes / comments must be backed by your own evaluation of the traffic. Do not submit reports about clients you have not actually observed on infrastructure you operate.

2. What you must not submit

Personal data (PII), credentials, secrets, or content from any third-party service in any free-text field.
Reports, votes, or comments motivated by spam, harassment, retaliation, or competitive interference.
Reports against IPs you know are search-engine crawlers, legitimate monitoring services, or your own infrastructure.
Mass-voting or sock-puppet voting from multiple installs you operate to manipulate the aggregate score. The hub's abuse detection flags this for review.
Anything illegal under the laws applicable to you.

3. Public visibility

The following fields are visible to anyone pulling the Community Bans feed or browsing it via the admin UI:

Your install's handle name (HN, e.g. "swift-otter-a3f7", or the display name you set instead), its country code (see §1), and every free-text reason / comment you attach.
Your vote kind (like / bad) and the parent submission you voted on. Votes are not anonymous; the HN is shown.
The 5-tier aggregate score assigned by the hub and the short reasoning string produced by the heuristic and/or AI judge.

Do not include anything you would not say in public. Do not assume votes are anonymous.

4. Licensing of your submissions, votes, and comments

You grant unmask.sh and every subscriber a perpetual, worldwide, royalty-free license to copy, store, display, aggregate, judge, and redistribute the data you submit, vote on, or comment on, for the sole purpose of operating and consuming the Community Bans service. You retain ownership of your own data.

5. How the hub uses submissions

Submissions are stored in a hub-side database keyed by the SHA-256 of an anonymous token (= no email, no account). Raw tokens are never persisted.
The hub aggregates by (IP, JA4) pair, folds in like / bad votes and whether the reporting installs have been flagged for abuse, and computes a 1-5 score. The hub may also ask an AI model for a second opinion; that request carries the reported IP and JA4 with the aggregate counts, and nothing that identifies the reporting installs (see the privacy policy).
Public-feed entries rotate out a fixed period after their most recent report, so old one-off reports do not keep showing up; see the privacy policy for the full retention model. Per-day salted hashes of reporting installs' IP addresses are scrubbed after 30 days, independent of any other retention, so the hub's database no longer ties a report to the IP address it came from.
The hub may reject submissions or revoke tokens that appear abusive (= mass false reports, sock-puppet voting, rate-limit violations).

6. Enforcement is on by default and defaults to a challenge

The browse feed includes both promoted entries (= a match kind and a score ≥ 3) and reports-only entries (= everything else, listed for context). Installs pull the feed every hour and enforce it by default. There is no local score threshold: which entries are enforced is decided on the hub, and each install sets what a match costs the visitor. The shipped defaults are:

subscribe_mode = fetch_apply (= pull and enforce). fetch pulls for browsing only; off stops pulling.
Action = pow_then_captcha (= a proof-of-work step, then a CAPTCHA). By default the feed never blocks anyone outright: real humans on flagged IPs still reach the site after passing the challenge. captcha_only, pow_only, and deny can be selected instead.

Search-engine crawlers and your own bypass IPs are exempt, and you can mute individual entries on your own install. Operators who change either default are responsible for the operational consequences (= false positives, escalated friction for legitimate users).

7. No warranty, no liability

The Community Bans service is provided as-is, with no warranty of any kind. unmask.sh is not liable for damages arising from feed accuracy, downtime, false positives, AI judge mistakes, or the operational decisions you make based on it. You are responsible for your own infrastructure and your own BAN list configuration.

8. Revocation, deletion, and corrections

You may stop submitting at any time (= flip submit_enabled off in admin). You may delete your own submissions, comments, and votes individually from the Bans page (= the inline × buttons in the detail panel). To remove every row associated with your token at once, use the removal request form with the HN your install displays.

Reported parties: if a banned IP is yours and should not be on the feed, use the removal request form with the IP. We will remove the entry.

9. Changes

We may update these terms. Material changes will require re-acceptance in admin before further submissions are accepted.

v1.1 — 2026-09-30: Corrected the description of the software to match what it does: subscribing (pulling and enforcing the feed) is on by default; publish_country defaults to ON; the setting is subscribe_mode; there is no local score threshold, and the default action is pow_then_captcha; the optional AI second opinion receives the reported IP and JA4; abuse detection flags vote manipulation for review rather than rate-limiting it. Your obligations under these terms are unchanged.
v1.0 — 2026-05-27: First published.

This document is not legal advice. unmask is an open-source project; consult counsel before relying on these terms for compliance in your jurisdiction.