Acceptable

Mullvad VPN

Commercial VPNSwedenRank 2 of 68

Paying monthly

$5.69 per month

Paying yearly

$69.60 billed once a year

Coverage

50 countries with an exit

Free tier

No paid plans only

60 out of 100

Security

64 out of 100

Usability

Visit site

Offered countries that sit under a logging mandateC09_mandate_countries

Turkey

Country of the ultimate beneficial ownerM11_owner_jurisdiction

Owner group cluster keyM15_owner_group

independent

Date the current owner took controlM16_control_since

not applicable

Role of this brand inside the groupM17_brand_role

principal

Infrastructure shared with cluster siblingsM18_shared_infrastructure

separate

Total published auditsV10_audit_count

Service class

Commercial VPN — security leads

Security score

60 of 100 · Acceptable

Usability score

64 of 100 · state known · 75% established · 79.4% of the weight read

Legal entity

Mullvad VPN AB, Box 53049, 400 14 Gothenburg, Sweden

Jurisdiction

Sweden

Ultimate owner

Amagicom AB, 100% owned by founders Fredrik Stromberg and Daniel Berntsson

Exposure

Medium · dominant channel: retention

Retention

Nothing that links traffic to a person

Latest audit

Assured Security Consultants (GotaTun code review) and Leviathan Security Group (Android Mobile App Profile assessment) — two engagements, both reports dated 2026-02-17 · 17 Feb 2026 · Repeated, at no fixed interval · Yes, and the report was read

Countries with exits

Protocols

WireGuard

Works under active censorship

Claimed to work

Weakest protocol still offered

not applicable

Own DNS inside the tunnel

Yes, on the provider's word

Platforms

Android iOS Linux macOS Windows

Simultaneous devices

Kill switch

unknown

Split tunnelling

Available

Signing up

No account needed

Free tier

No free tier

P2P allowed

unknown

Major streaming services reachable

unknown

Price, monthly

$5.69 per month

Price, yearly

$69.60 a year at renewal, billed once

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

54 7 left unknown, 6 not applicable

Completeness

87% the share the model could weigh

Citations

74 across 25 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 · 9 hereB weighs 0.85 · 61 hereC weighs 0.70 · 4 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 audit cadence. Mullvad publishes an audit index in the app repository and it is honest about cadence: the index states outright that the main app audits happen every two years, with extras in between, which is why regularity is recorded as repeated at no fixed interval rather than as an annual series — no annual cadence is claimed by anyone.

Two reports, one date. Two separate engagements share 2026-02-17 and the record names both: Assured's code review of GotaTun, the userspace WireGuard implementation (no critical, high or medium issues; two low and nine notes), and Leviathan's Mobile App Profile assessment of the Android app, whose compliance report gives February 17, 2026 as its date of evaluation and AL2 as its assurance level. One date, two reports, no contradiction.

the rest of what is open — 225 more words

Why the count reads eight. The audit count is eight because the GotaTun review sits outside the app repository's own list. Before them, a 30 person-day X41 D-Sec review across all five platforms found six vulnerabilities — three high, two medium, one low, none critical.

What no engagement covers. What none of these engagements covers is the server fleet: every one of them examines the app or the tunnel protocol, which is why fleet coverage is recorded as not applicable rather than as a sample. The no-logs claim on the infrastructure side therefore rests on the April 2023 search warrant — Swedish police came for customer data and left without it — and not on any audit.

Reproducible builds, Android only. Reproducible builds are Android-only, from version 2025.2 onward, so the field stays partial; the desktop and iOS builds are not verifiable bit-for-bit. Turkey is the one mandate country on the offered list and the hosting behind it is not established.

Nothing published about demands. Nothing on record rests on anything the provider publishes about demands it receives: there is no transparency report and no warrant canary, so the April 2023 seizure carries that ground alone. Kill-switch coverage and P2P policy are unestablished; the latest audit is a protocol code review, not a no-logs engagement, so V11 is silent and retention leans on the declaration.

On the record as a whole.

The April 2023 search warrant is the strongest single item in this record: police attended the office and left with nothing, which is practice evidence rather than paper and ages far more slowly than an audit. Pricing is a flat 5 EUR/month on every term; U04_price_monthly carries 5.80 (the USD equivalent Mullvad states on its own page); U04b_price_yearly is 69.60, the full year at that rate.

RetentionDoes a record exist that links traffic to a person, and who checked?13 fields read · 2 not applicable · 21.8 points of pressure×0.782

Composed rather than added: a prior of 0.55 on the silent branch, moved by what the paper says (0.0687) and by what the practice shows (0.608), at credence 0.6389 from the audit and 0.855 from how the provider engaged, against a severity of 0.03 for what is actually retained.

Retention linkable to a person as declaredmaterialC01_linkable_retention

none_linkable

Bmullvad.net13 Aug 2026

Report a problem with C01_linkable_retention

Persistent device identifier the provider holdsC12_device_identifier

none

Bplay.google.com15 Aug 2026

Report a problem with C12_device_identifier

At least one independent auditmaterialV01_audit_exists

yes_verified

Bgithub.com17 Feb 2026

Report

Computed

×0.601 of a clean record completeness 87%

Where the drop comes from: Retention. It is the channel this record's risk is concentrated in — on its own it takes a clean record to ×0.782, and its bar above is filled to that.

Exposure is recorded as medium and the integrity status as clean.

This record sits above the register's average on all 8 channels. The narrowest margin is Disclosure, ×1.000 against ×0.970.

Published score

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.

Specification

Carried, never scored

Countries with an exit

Compared with

What changed

If this one is not right

Mullvad VPN

Proton VPN

IVPN

AirVPN

Where the security score comes from