Where the security score comes from
Eight channels, each asking one question of the published record and reading a named set of fields to answer it. They multiply rather than add — the factor beside a channel is what is left after the pressure it found, and open a channel to see every field it read, as it is stored. How the model works
Security fields
67 55 of them feed a channel
Established
57 9 left unknown, 1 not applicable
Completeness
84% the share the model could weigh
Citations
91 across 32 documents
Every value carries what it rests on, and the model weighs the four levels differently. Weight, then how many citations on this record sit at that level:
A weighs 1.00 · 15 hereB weighs 0.85 · 64 hereC weighs 0.70 · 11 hereD weighs 0.35 · 1 here
The weights are methodology.json’s own; a claim seen in the artefact itself counts for more than the same claim reported second-hand. The date at the end of a row is the day that value was last read.
A channel does not price every field it reads. The ones it did are marked along their edge and carry what they added; everything else was read and cost nothing. Only a channel that composes field by field can mark a field at all — the rest price themselves by the parameters printed above their rows.
Still open on this record.
The pentest, and the one not done. The pentest field is backed by Cure53's October-November 2024 review of the Lightway protocol, which found an unauthenticated-data-fragment denial of service rated High alongside four lesser items. ExpressVPN's trust centre lists a long series of such engagements, so the cadence is real; what it does not contain is an infrastructure pentest of the server fleet separate from the KPMG no-logs engagement.
The two prices, corrected. The pricing page now sells three separate plan lengths — 1-month, 12-month and 2-year — rather than the 24-28-month-only structure read on 2026-08-29. U04_price_monthly=14.99 is the Basic 1-month plan, autorenewing at the same $14.99. U04b_price_yearly=59.85 is what the Basic 12-month plan (billed as 12 months plus a 3-month bonus) actually charges for its first term; it autorenews at $99.95/year afterward, the figure the previous record had mistakenly stored as the yearly field itself.
the rest of what is open — 363 more words
The transparency report. P05/P06/P08/P09 now rest on ExpressVPN's biannual transparency report (Jul–Dec 2025: 155 government/law-enforcement requests, 0 disclosures, 3 warrants listed separately) reached via the Trust Center — the earlier run missed both.
The canary, and the rest of the block. No ExpressVPN warrant canary could be found and no primary statement denies one, so P07 stays unknown. P02/P03=none_known rest on the transparency series and the privacy policy's litigation stance; P11=none_known on the policy's statement that all demands run through BVI courts. P04=withdrew_physical is the June 2022 India exit.
What is never published. Subscriber count (M08) and total server count (U03) are not published. Server hardware ownership (C08: owned vs leased) and app-level device identifiers (C12) could not be established from any page checked.
106 countries against a claim of 113. U02_server_countries reflects the 105 country-landing pages in ExpressVPN's own sitemap plus China (confirmed 'Available - Virtual Locations' on the live /vpn-server page but lacking a dedicated landing page) — 106 codes total, against the site's own marketing claim of '113 Countries'; the live server-location page is a very large per-protocol table that was not exhaustively parsed, so the gap between 106 and 113 is reported rather than resolved.
An audit finding between two categories. V11_audit_findings is recorded as session_metadata rather than none_linkable or aggregates_only: the 2025 KPMG report describes a structured per-connection record (country, ISP, date, protocol, data volume, a stable salted-hash pseudonym) that is more than a pure aggregate but lacks IP and precise time — a genuinely ambiguous case for this field's four middle categories.
Not checked this pass. Whether ExpressVPN's coverage exists outside affiliate sites (X08), its affiliate payout tier (X10), and whether owner-run review media flags the ownership tie (X11) were not checked in this pass.
P2P, looked for and absent. U12a_p2p_allowed stays unknown for a different reason than the rest: it was looked for and is not published. Neither sitemap (578 page and support URLs) nor the terms of service mentions P2P, BitTorrent or file sharing anywhere, while every other catalogued provider states its position in its own documents. A support article or a terms clause would close it; a review site saying it works would not.
On the record as a whole.
Express Technologies Ltd., incorporated in the British Virgin Islands since 2009, owned by Kape Technologies plc since the September 2021 acquisition ($936M, Kape's own about-us page). TrustedServer RAM-only architecture has been audited annually by KPMG (ISAE (UK) 3000 Type I, most recently 28 Feb 2025) since 2022 for privacy-policy compliance, plus PwC Switzerland (2019, TrustedServer) and Cure53 (2019 browser extensions; 2021 Lightway protocol, open-sourced).
The 2025 KPMG report was read directly (confidence A): the server writes one pseudonymous event per connection (date without time, salted/hashed username, country+ISP GeoIP, data volume, location group, protocol) but never logs IP address or exact timestamp; builds are described as reproducible across two independently operated build systems.
the rest of what is open — 318 more words
In 2017, Turkish authorities seized an ExpressVPN server investigating the assassination of Russian Ambassador Andrei Karlov and found no usable logs (TorrentFreak, which carries a sponsorship notice for ExpressVPN — noted, not excluded, since the facts are externally checkable and ExpressVPN's own statement page has since returned HTTP 410 with no Wayback Machine snapshot).
In September 2021 the DOJ's Deferred Prosecution Agreement with Daniel Gericke, ExpressVPN's Chief Information Officer, over his prior 2016-2018 role in the UAE's 'Project Raven' hacking operation became public; the primary DPA document itself never names ExpressVPN (it covers only his earlier UAE employment) — the ExpressVPN link comes from ExpressVPN's own and press statements. ExpressVPN kept Gericke employed and published a public defense of that decision.
C12 is resettable_ad_id on the provider's own prose rather than on a store category. The privacy policy's marketing paragraph states that an Installation ID and an XV User ID are always collected and that advertising identifiers — IDFA or the Android Advertising ID — follow on explicit consent, all of it reaching AppsFlyer; the Play data-safety form independently declares Device or other IDs for advertising or marketing.
The consent gate and a mitigation nobody else here offers — AppsFlyer-bound traffic routed through a dedicated ExpressVPN IP pool so attribution cannot correlate by IP — are real and are recorded here, because the floor prices what can be held, not how well it is explained.
U07_protocols carries the whole picker: Lightway UDP and TCP as vendor_open, beside WireGuard, OpenVPN UDP/TCP and IKEv2. Lightway is ExpressVPN's own protocol and its source is published with independent review behind it (Cure53, 2021 and 2024), which is the distinction vendor_open exists for — proprietary prices the impossibility of telling whether a transport is weak, and here it can be told.
IntegrityHas it been caught saying one thing and doing another?16 fields read · 5 left unknown · 63.3 points of pressure×0.367
What this channel priced on this record:Monetisation 0.2Bought coverage 0.1Incident handling 0.2Implausible report 0.15Owner group history 0.25
Transparency reportalso read by ignorance, silenceP05_transparency_report
fresh_with_numbers
Bexpressvpn.com
Computed
×0.144 of a clean record completeness 84%
Where the drop comes from: Integrity. It is the channel this record's risk is concentrated in — on its own it takes a clean record to ×0.367, and its bar above is filled to that.
Exposure is recorded as medium and the integrity status as clean.
Against the rest of the register, this record sits furthest above on Jurisdiction — ×1.000 where the register averages ×0.680 — and furthest below on Integrity, ×0.367 against ×0.658.
Every row above is one field of the published record, and every one of them can be challenged: every row carries Report, or raise something about the record as a whole. A correction with a source behind it is made the same day, and if it moves the score it moves the rank.
