Every juror is named. Every limitation is visible.

Public-IP lookup remains fail-closed until the deployed IPBot build has an approved, unexpired field-level public-display profile. When active, IPJury requests only sanitized network, routing, classification, direct-evidence, and geolocation fields and omits global scores, raw sources, and internal rule identifiers.

The LOCATION row labelled IP2Location is IP2Location LITE DB11 data served through IPBot's audited build; IPBot is the delivery path, not a second estimate, so the database appears once in the consensus and once in the matrix. Region and city are estimates about the containing block. Attribution: IPJury uses the IP2Location LITE database for IP geolocation.

Looks up a PTR hostname derived from the target IP. PTR is operator-controlled and never treated as identity proof.

Optional country and ASN evidence. When enabled, the target IP is disclosed to IPinfo and required attribution appears in the record.

Bright Data's geo.brdtest.com endpoint answers with a country, region and city for the containing block. It publishes no terms and does not declare the database behind it, so it is shown on the owner's decision of 2026-09-15 as a context row only: it never votes in the country consensus, never moves the band and never reaches the score, and its agreement with another row is not counted as corroboration. IPv6 targets are never sent, because the endpoint's IPv6 answers named the wrong network in testing. When enabled, the target IPv4 address is disclosed to Bright Data. A permission request is drafted; the row is removed the same day on any objection. Attribution: geolocation context by Bright Data.

A locally held monthly edition of the DB-IP City Lite database, read on the first-party ledger host so the target IP is never sent to DB-IP. Contributes its own country, region, and city row to the location consensus, always scoped to the containing block. Attribution: IP Geolocation by DB-IP.

A locally held GeoLite2 Country edition, refreshed twice weekly under the GeoLite EULA's update duty and read on the first-party ledger host — the target IP is never sent to MaxMind. Contributes a country-level row to the location consensus. City-level GeoLite data is not used: MaxMind declines it for this account under the US bulk sensitive data transfer rules, and that refusal is respected rather than sourced around. This product includes GeoLite Data created by MaxMind, available from https://www.maxmind.com.

A locally held GeoLite2 ASN edition joining the routing axis beside IPtoASN, so the announced-origin answer is corroborated or disputed across two independent tables. Scoped to the containing block, never the single address. This product includes GeoLite Data created by MaxMind, available from https://www.maxmind.com.

The published botnet command-and-control blocklist, imported on a schedule and held with its editions so the record can say when IPJury first saw a listing. Membership is direct adverse evidence on the abuse axis; absence means this source did not observe the address — never that it is clean. Data by Feodo Tracker (abuse.ch), released under CC0.

The published exit list is imported on a schedule and held with the edition it came from, so the record can say when IPJury first saw an address on it and when its membership last changed. Membership is a network role on the anonymity axis and never an abuse finding. Absence means this source did not list the address — not that the address is not a Tor exit — and once an edition ages past its refresh window the juror abstains instead of continuing to assert.

Mullvad publishes the list of addresses its relays are reached on, and it is imported on a schedule and held with the edition it came from, so the record can say when IPJury first saw an address on it and when that listing ended. The published field is the address a relay listens on, so membership never asserts where traffic leaves. Running a VPN relay is a network role on the anonymity axis and never an abuse finding. Absence means this source did not list the address — not that the address is not a VPN — and once an edition ages past its refresh window the juror abstains instead of continuing to assert.

The registry's own designation for an address block, with its three-state answer on global reachability preserved rather than flattened. It states what a block is set aside for; it says nothing about the behaviour of anything inside it.

Which network announces the block an address sits in, held as the published start–end range rather than reshaped into prefixes the source never named. Because the ledger keeps editions, the record can also say how many times that announcement has changed since IPJury first saw it. An announcing network is who routes the block — not who uses the address, and not what they do with it.

Each registry's daily delegated-statistics file, read from that registry's own server and held with its editions, gives the REGISTRY row a source of its own: the country a block was first allocated or assigned to, on what date, by which registry. That is a registration fact about the block and never a statement of where the address is in use, so it is shown beside the location estimates and never counted among them. Blocks a registry lists as available or reserved are recorded as not allocated, never as a country. The RIPE NCC file is not read — its site-wide terms limit reuse to non-commercial purposes — so addresses RIPE NCC administers abstain on this row rather than borrowing another reading.