Head to head
The gap
The security score is not a total of good things. It starts at 100 and eight channels each take a share of what is left, so one channel can settle a record on its own. Both records are drawn here on the same scale — the model’s own cap of 0.9 — rather than on a scale fitted to this pair, which would make every comparison look equally dramatic.
Line by line
Every field the contract publishes about a provider, in the contract’s own groups, including the rows where the two agree — a table that printed only differences would invent a disagreement, and one that printed only the interesting rows would be us choosing what is interesting. Open a group to read its rows; each one says how many it holds and how many of them the two records answer differently.
A green cell means the methodology itself prices that value below the other one, not that we prefer it. Rows the model does not price carry no mark at all: price is never marked, because nothing in the model says cheaper is safer, and neither is coverage, platform count or devices. Where a value was never established the cell says so instead of showing a zero.
What actually happened
Other pairs
Pairs within the top ten get a page — 80 comparisons in all. Records below that are not paired here: a comparison earns a page where the choice between the two is a real one, and that choice is made near the top of the register.
Security
Usability
Rank 9 of 68 in the register
Romania · Commercial VPN
Exits in 91 countries
Security
Usability
Rank 6 of 68 in the register
United States · Commercial VPN
Exits in 91 countries
No demand on the public record
1 documented demand, none of which produced data
Open questions
Still open on this record.
Owner-group history. Kape Technologies, CyberGhost's parent since March 2017, traces back to Crossrider, an ad-injection SDK platform whose services were widely abused by adware and malware distributors before the 2016 pivot to VPNs. No documented lie or claim gap ties this history to CyberGhost's own conduct, but the adverse record is why this provider is catalogued at all rather than by a neutral selection rule.
Scale and reach are undocumented. No CyberGhost-specific user count could be sourced from the provider's own pages beyond "millions of customers," so M08_user_scale stands unknown, and no clean 12-month plan exists to price against U04b_price_yearly — CyberGhost sells only 1-month, 6-month and 26-month terms.
Client and build transparency are unexamined. No official CyberGhost source-code repository, reproducible-build attestation or standalone application pentest (Cure53's public index has no entry for CyberGhost or Kape) could be found, leaving V12, V13 and V14 unknown rather than a finding either way.
A disputed breach claim was not independently resolvable. A September 2024 third-party report alleged leaked CyberGhost account credentials; CyberGhost's own response denies a service-side compromise and attributes the credentials to user-side malware, but the underlying report could not be read in this session, so X09_incident_handling stays unknown rather than taking either side's account.
Two of CyberGhost's own pages disagree on its street address: the imprint names 70-72 Dionisie Lupu street, the privacy policy names 68 Polonă street, both in the same Bucharest district and both current pages rather than one being an archived leftover. Neither was treated as authoritative over the other, and the discrepancy is left for a future refresh to resolve rather than guessed at here.
Still open on this record.
What the report is. The security score rests on a Deloitte Audit SRL limited assurance report under ISAE 3000 (Revised), signed 19 December 2025 for the configuration as of 31 October 2025. Read what that is: the conclusion is negative assurance — "nothing has come to our attention" — against a description PIA's own management wrote, and the report states it provides no assurance for any other period.
Who the report is for. The auditor also restricts the report to PIA management and disclaims responsibility to any other reader, so the strongest document in this record is one whose author declines a duty of care toward the person reading it here. Procedures were performed on a sample basis, which is why fleet coverage is recorded as a sample.
What it did not examine. The report covers infrastructure only. C07a_app_sdk and C12_device_identifier are app-side facts it never examined; they now cite the privacy policy instead, which describes analytics for website visitors and opt-in diagnostic reports carrying platform and app version. Reading either as a statement about SDKs compiled into the app is the weakest link in this record, and the open-source clients are where it would actually be settled.
The 2016 subpoena rests on PIA's own word, not the court file. The only detailed account of the McWaters case (S.D. Fla., 9:16-cr-80043) sits on an affiliate site this record cannot cite; CourtListener's docket for that case names no subpoena, no provider and no outcome. P03_court_record and P12_case_history now stand on PIA's own transparency-report statement that it was subpoenaed in 2016-2017 and produced nothing — a B-level self-declaration, not a verified court record.
Recent changes
No change logged on this record
Notes
On the record as a whole.
Deloitte's third no-logs assurance report (the only one of three published as a full downloadable PDF rather than gated behind login or an email request to Deloitte) is a limited assurance, ISAE 3000 (Revised) engagement dated as of 31 October 2025 and signed 19 December 2025; it examines CyberGhost's description of its own no-logging controls on a sample basis, not the full fleet, and its conclusion is negative assurance ("nothing has come to our attention"), not a positive finding of no logs.
The privacy policy's non-personal-data list names the Google Advertising ID and gyro-sensor data under website-style language that only a mobile app can actually read.
On the record as a whole.
First fill 2026-08-26, refreshed 2026-09-12 on P03_court_record, P12_case_history and V12_client_open_source, refreshed again 2026-09-13 on M08_user_scale, P02_disclosure_history, P03_court_record, P05_transparency_report, P06_report_plausibility, P08_requests_received, P09_data_provided and P12_case_history, and refreshed again 2026-09-13 on U04_price_monthly and U04b_price_yearly.
The 2026-09-13 pass found PIA had published its Q1 2026 and Q2 2026 transparency reports since the record's last look at Q4 2025; the record now cites the Q2 2026 post (19 requests received, 0 resulting in data provided, covering April-June 2026, published 2026-08-06) as the freshest, with the same subpoenas/warrants/other/foreign breakdown structure as before.
M08_user_scale, P02_disclosure_history, P03_court_record and P12_case_history were re-read against the same sources and stand unchanged; the main transparency-report page still names no case beyond the 2016-2017 subpoenas already on file. The pricing re-read against the same buy-vpn-online page found both figures unchanged: month-to-month $11.95, one-year plan $47.88 billed for the first year and autorenewing at the same $47.88 with no intro/renewal gap, consistent with PIA's flat-rate pricing noted elsewhere in the catalogue.
Sourced from privateinternetaccess.com's own pages and blog, the two Deloitte ISAE 3000 (Revised) 'Independent Limited Assurance Report' PDFs (April 2024 and December 2025, both opened and read directly), Kape Technologies' own site, the Isle of Man/Companies House-sourced reporting on Kape's 2023 take-private by Unikmind Holdings (Teddy Sagi), and the official pia-foss GitHub organisation for the open-source clients.
The public CourtListener docket for United States v. McWaters (9:16-cr-80043, S.D. Fla.) was opened but names no subpoena, VPN provider or outcome; the affiliate writeup (vpncompare.co.uk) that describes that subpoena in detail was dropped as a forbidden source, and TorrentFreak's own coverage returned HTTP 403 to every fetch attempt this run, curl included. P03/P12 now rest on PIA's own transparency-report self-declaration instead.
Wikipedia was dropped from V12 in favour of the GitHub organisation itself. No affiliate or 'best VPN' aggregator site was used as a factual source anywhere in this record, even though several turned up repeatedly in search results with plausible-looking numbers (server counts, user counts, audit dates) that could not be cross-checked against a primary page and were discarded rather than used.
13 Sep 20261 changes
Why this provider is in the catalogueM14_inclusion_basis
hand_picked, adverse_record changed to adverse_record
12 Sep 20263 changes
Chronology of documented enforcement eventsP12_case_history
{"date":"2016-03","kind":"subpoena","outcome":"nothing_produced","source_url":"https://www.vpncompare.co.uk/private-internet-access-no-log-claim-put-to-the-test/"} changed to {"date":"2016-03","kind":"subpoena","outcome":"nothing_produced","source_url":"https://www.privateinternetaccess.com/transparency-report"}
Security scoresecurity.score
27 changed to 25
Requests received in periodP08_requests_received
30 changed to 19
30 Aug 20261 changes
Weakest protocol still offeredC13_weakest_protocol
wireguard changed to not on the record
| ChannelPressure taken from what was left | CyberGhost VPNDominant channel: Retention0 to 0.9, longer is worse | Private Internet AccessDominant channel: Integrity0 to 0.9, longer is worse |
|---|---|---|
| IntegrityHas it been caught saying one thing and doing another? | 0.325 | 0.459 |
| RetentionDoes a record exist that links traffic to a person, and who checked? | 0.450 | 0.188 |
| JurisdictionCan someone compel what exists, without the provider getting a say? | 0.136 | 0.272 |
| IgnoranceHow much of this did nobody establish? | 0.181 | 0.071 |
| HygieneWhat identifying material is collected in passing, and can the client be checked at all? | 0.150 | 0.101 |
| ArchitectureWho operates the exit, and what does it do to the traffic? | 0.050 | 0.050 |
| DisclosureHas it already handed something over? | 0.000 | 0.000 |
| SilenceTen years, millions of users and nothing on record? | 0.000 | 0.000 |
| What is left of 100The channels multiply: each one takes a share of what the one before it left standing | 21 | 25 |
By circumstance
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Torrenting | Allowed yes | Allowed yes |
| Streaming | Reachable yes | Reachable yes |
| Devices at once | 7 | unlimited |
| Under censorship | — not established | Claimed to work claimed |
| Exits in mandate countries | 7 United Arab Emirates, China, India, Iran, Kazakhstan, Turkey, Vietnam | 6 United Arab Emirates, China, India, Kazakhstan, Turkey, Vietnam |
| Hosting in those countries | Physical and virtual both mixed | Physical and virtual both mixed |
| Company under a mandate | No such mandate no | No such mandate no |
| Secret orders | — not established | National security letters or their equivalent nsl_equivalent |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Security | 21 insufficient · rank 9 of 68 | 25 insufficient · rank 6 of 68 |
| Usability | 69 out of 100 | 73 out of 100 |
| Record completeness | 76% usability 92% | 89% usability 100% |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Who operates the exit | Run by the provider itself provider_operated | Run by the provider itself provider_operated |
| Retention as declared | Aggregate counts only aggregates_onlyBcyberghostvpn.com · | Nothing that links traffic to a person none_linkableBprivateinternetaccess.com · |
| RAM-only diskless servers | Yes, checked by someone independent yes_verified | Yes, checked by someone independent yes_verified |
| IP allocation model | Dedicated address offered as an option dedicated_option | Dedicated address offered as an option dedicated_option |
| Signup without email | No, an email address is required no | No, an email address is required no |
| Anonymous payment options | Cryptocurrency through a processor | Cryptocurrency through a processor |
| Own DNS inside the tunnel | — not establishedno source published | Yes, on the provider's word yes_declaredAassets-cms.privateinternetaccess.com · |
| Worst third-party SDK category in the app | Install attribution attributionBcyberghostvpn.com · | Analytics analyticsBprivateinternetaccess.com · |
| Worst third-party tracker category on the site | Analytics analytics | Analytics analytics |
| Who owns the hardware | — not establishedno source published | Owned by the provider ownedBprivateinternetaccess.com · |
| Offered countries that sit under a logging mandate | 7 United Arab Emirates, China, India, Iran, Kazakhstan, Turkey, VietnamBcyberghostvpn.com · | 6 United Arab Emirates, China, India, Kazakhstan, Turkey, VietnamBserverlist.piaservers.net · 2 sources · |
| Hosting in the mandate countries it offers | Physical and virtual both mixed | Physical and virtual both mixed |
| Incorporation country mandates retention | No such mandate no | No such mandate no |
| Beneficial owner sits under a retention regime | No such regime no | No such regime no |
| Persistent device identifier | A resettable advertising identifier resettable_ad_idBcyberghostvpn.com · | An identifier scoped to the app app_scoped_idBprivateinternetaccess.com · |
| Weakest protocol still offered | — the question does not arise | — the question does not arise |
| What the exit does to passing traffic | — not established | — not established |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| At least one independent audit | Yes, and the report was read yes_verified | Yes, and the report was read yes_verified |
| How the report is published | The full report is published full_pdf | The full report is published full_pdf |
| Subject of the audit | The no-logs claim itselfAdownload.cyberghostvpn.com · | The no-logs claim itself · The infrastructureAassets-cms.privateinternetaccess.com · |
| Server fleet coverage | A sample of servers sample | A sample of servers sample |
| Auditor name | Deloitte Audit S.R.L. (Deloitte Audit Romania)Adownload.cyberghostvpn.com · |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Server seizure by law enforcement | No seizure on record none_known | No seizure on record none_known |
| Actual disclosure history | No disclosure on record none_known | No disclosure on record none_known |
| Court record of compelled production | No court record none_knownno source published · | Told a court it had nothing to produce unable_to_produceBprivateinternetaccess.com · |
| Response to a national logging mandate | — not establishedno source published | Pulled its servers out rather than comply withdrew_physicalBprivateinternetaccess.com · |
| Transparency report | Current, with request numbers fresh_with_numbers |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Documented gap between claim and behaviour | No documented lie noneno source published · | — not establishedno source published |
| Marketing against audit findings | No gap on record noneno source published · | — not establishedno source published |
| Ownership structure transparency | Ownership is stated and checkable clearBcyberghostvpn.com · 2 sources · | Ownership is partly stated partialBprivateinternetaccess.com · 2 sources · |
| Adverse events on the owner group record | A history of adware or malware adware_malware_history | A history of adware or malware adware_malware_history |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Legal entity | CyberGhost S.R.L.Bcyberghostvpn.com · | PIA Private Internet Access, Inc., 5555 DTC Parkway, Suite 360, Greenwood Village, CO 80111, USAAassets-cms.privateinternetaccess.com · |
| Where the company sits | Romania ROBcyberghostvpn.com · | United States USAassets-cms.privateinternetaccess.com · 2 sources · |
| Year the service launched | 2011Bcyberghostvpn.com · | 2010 |
| CyberGhost VPN | Private Internet Access | |
|---|---|---|
| Platforms with a client | 9 Windows, macOS, Linux, Android, iOS, Android TV, Fire TV, Routers, Browser extensionBsupport.cyberghostvpn.com · 2 sources · | 9 Windows, macOS, Linux, iOS, Android, Routers, Android TV, Apple TV, Browser extensionBprivateinternetaccess.com · 2 sources · |
| Countries with an exit | 91 countriesBcyberghostvpn.com · | 91 countriesBserverlist.piaservers.net · 2 sources · |
| Paying monthly | $12.99 per monthBcyberghostvpn.com · |
CyberGhost VPN and Private Internet Access were assessed by the HushFleet team and both placed in the Insufficient band: CyberGhost VPN scores 21/100 for security, Private Internet Access 25/100. They share an independent audit, one owner group and the same three protocols. The 4-point gap opens widest on Retention — “Does a record exist that links traffic to a person, and who checked?” — taking 45% from CyberGhost VPN and 19% from Private Internet Access.
| Deloitte Audit SRL (Romania)Aassets-cms.privateinternetaccess.com · |
| Auditor category | One of the Big Four big4 | One of the Big Four big4 |
|---|
| Assurance standard | ISAE 3000 isae3000 | ISAE 3000 isae3000 |
|---|
| Date of latest report | 2025-12-19 | 2025-12-19 |
|---|
| Audit cadence | Repeated, at no fixed interval repeated_irregular | Repeated, at no fixed interval repeated_irregular |
|---|
| Total published audits | 3 | 3 |
|---|
| Retention as the report states it | Aggregate counts only aggregates_onlyAdownload.cyberghostvpn.com · | Nothing that links traffic to a person none_linkableAassets-cms.privateinternetaccess.com · |
|---|
| Client source code published | — not establishedno source published | Every client allBgithub.com · 2 sources · |
|---|
| Reproducible builds | — not established | — not established |
|---|
| Separate pentest | — not established | — not established |
|---|
| Strength of the assurance opinion | A limited-assurance conclusion limited | A limited-assurance conclusion limited |
|---|
Current, with request numbers fresh_with_numbers |
| Reported numbers plausible for this scale | The numbers are plausible at this scale plausible | The numbers are plausible at this scale plausible |
|---|
| Warrant canary discipline | — not establishedno source published | Never had one noneBprivateinternetaccess.com · |
|---|
| Requests received | 11Bcyberghostvpn.com · | 19Bprivateinternetaccess.com · |
|---|
| Requests that produced data | 0 | 0 |
|---|
| Secret-order regime | — not establishedno source published | National security letters or their equivalent nsl_equivalentCeff.org · |
|---|
| Chronology of documented enforcement events | 0 No demand on the public recordBcyberghostvpn.com · | 1 documented demandBprivateinternetaccess.com · |
|---|
| How the service is paid for | Paid subscriptions only paid_only | Paid subscriptions only paid_only |
|---|
| Claims of 100% anonymity or military-grade crypto | Makes no such claim noBcyberghostvpn.com · | Claims total anonymity or military-grade encryption yesBprivateinternetaccess.com · 2 sources · |
|---|
| Aggressive lifetime subscriptions | Does not no | Does not no |
|---|
| Coverage exists only on affiliate sites | — not establishedno source published | Covered outside affiliate sites too noCtechradar.com · |
|---|
| Security incident and how it was disclosed | — not establishedno source published | No incident on record none_knownCprivateinternetaccess.com · |
|---|
| Affiliate programme and payout tier | A standard payout standard | A standard payout standard |
|---|
| Owner-run review media discloses the tie | — not establishedno source published | — the question does not ariseno source published |
|---|
| App store declaration | — not establishedno source published | The store form matches the published policy matches_policyBapps.apple.com · 3 sources · |
|---|
| User base scale (mass >=10M, mid >=1M, niche <1M) | — not establishedno source published | Mid: a million users or more midBinvestegate.co.uk · |
|---|
| Ultimate beneficial owner as written | Teddy Sagi, sole owner of Unikmind Holdings Limited (Isle of Man), which holds Kape Technologies plc — CyberGhost's parentBunikmind-holdings.com · 2 sources · | Teddy Sagi, via Unikmind Holdings Limited (Isle of Man), sole shareholder of Kape Technologies plc since the May 2023 delisting; Kape owns PIA Private Internet Access, Inc.Cen.wikipedia.org · 2 sources · |
|---|
| Other businesses of the owner group | Antivirus · A portfolio of other apps · A review directoryBkape.com · | Antivirus · A review directory · Adtech or adwareCen.wikipedia.org · 2 sources · |
|---|
| Country of the ultimate beneficial owner | IM | IM |
|---|
| Owner group | kape | kape |
|---|
| Date the current owner took control | 2017-03Bkape.com · | 2019-12Bkape.com · |
|---|
| Role of this brand inside the group | A brand the group acquired acquired_brandBkape.com · | A secondary brand secondary_brandCkape.com · |
|---|
| Infrastructure shared with cluster siblings | — not established | — not established |
|---|
| $11.95 per monthBprivateinternetaccess.com · |
| Paying yearly | — not establishedno source published | $47.88 a year at renewal, billed onceBprivateinternetaccess.com · |
|---|
| Usable free tier | No free tier none | No free tier none |
|---|
| Works under active censorship | — not establishedno source published | Claimed to work claimedBprivateinternetaccess.com · |
|---|
| Protocols offered | 3 WireGuard, OpenVPN, IKEv2 | 3 WireGuard, OpenVPN, IKEv2 |
|---|
| Kill switch coverage | On some platforms partialBcyberghostvpn.com · 2 sources · | On every platform all_platformsBprivateinternetaccess.com · |
|---|
| Split tunneling available | Available yes | Available yes |
|---|
| Simultaneous connections | 7Bcyberghostvpn.com · | unlimitedBprivateinternetaccess.com · 2 sources · |
|---|
| Friction to start using it | An email address only email_only | An email address only email_only |
|---|
| P2P allowed | Allowed yes | Allowed yes |
|---|
| Major streaming services reachable | Reachable yes | Reachable yes |
|---|
