A logging policy is a promise. The client is a fact.

Every VPN says roughly the same sentence about logs. It is the cheapest sentence in the industry to write and the most expensive one to verify, because from where you are sitting it cannot be verified at all. You are reading a claim about the internal behaviour of a computer you have never touched, published by the people who run it.

I run a small VPN and I want to be blunt about this: my logging policy is a promise, the same as everyone else's. What separates providers is not who writes the sentence more confidently. It is how much of their system you can check without taking anyone's word for it.

The record is not reassuring

This is not a hypothetical worry. The pattern has repeated for a decade.

In October 2017, US court documents in the Ryan Lin cyberstalking case showed that PureVPN had supplied records that helped investigators tie activity to a specific person, despite the company's marketing at the time. TorrentFreak covered the filings in detail.

In 2018 it came out that IPVanish, under its previous owner Highwinds, had handed over user records to Homeland Security in a 2016 case, again while advertising that it kept none.

In July 2020, security researchers found around 1.2TB of user records sitting on an unsecured Elasticsearch server, belonging to a family of related no-cost VPN apps that all advertised themselves as keeping nothing. Comparitech published the analysis. The records included plaintext passwords, IP addresses and connection timestamps.

Go back further and it gets worse. In 2015 Hola was found to be reselling its users' idle bandwidth through a sister company, which meant strangers' traffic was exiting through the machines of people who thought they had installed a privacy tool. Facebook's Onavo Protect, marketed as a VPN, was in effect a market research probe on its users' phones, and Apple pulled it from the App Store in 2018.

Not one of those companies had a policy page saying "we collect your connection logs and may hand them over." They all said the opposite. The page was never the problem. The page was the product.

Third-party audits help, and they are not proof either

The industry's answer to this is the audit. It is a real improvement and I am not going to sneer at it. An auditor who is allowed to look at running infrastructure will catch sloppiness that marketing copy hides.

But be clear about what an audit is: a snapshot, taken on a date, of a system the provider controls, under a scope the provider negotiated and paid for. It says something true about a Tuesday in March. It says nothing enforceable about the following Tuesday, and audit reports are frequently summarised in a blog post rather than published in full.

Pangea has not been audited by a third party. We are small and it costs money we do not have yet. I would rather write that sentence than let you infer otherwise from a badge on a landing page.

What you can check without trusting me

Here is the part that is not a promise.

The Pangea desktop app is open source under GPLv3. Not "source available", not a marketing repo with the interesting parts missing. You can read the code that runs on your machine, and if you think I am wrong about something in it, you can say so in public with a line number attached.

Your private key is generated on your device. That is checkable in the source, and it is checkable on the wire: watch what the app sends during setup. A key that never leaves your laptop is not a key I can be compelled to hand over, because I do not have it. That is a structural fact about where the maths happens, not a policy I could quietly change in an update without anyone noticing.

The tunnel is WireGuard with 256-bit encryption (ChaCha20). WireGuard's whole design argument is that it is small enough to read. The reference implementation is a few thousand lines, which is why cryptographers have actually reviewed it, unlike the sprawling stacks it replaced.

So the honest split looks like this. Everything that happens on your computer, you can verify yourself, today, with no permission from me. Everything that happens on the server side rests on trust, mine and everybody else's, and anyone who tells you different is selling you a feeling.

The structural answer to a trust problem

If server-side behaviour cannot be verified from outside, the useful question changes. It stops being "do they keep records" and becomes "what could a record possibly be attached to."

You can pay us in crypto, including Monero. That path takes no email address and no KYC, and it never touches a payment card. If you use it, we do not know who you are, and nothing we might or might not keep on a server can be joined up to your name, because your name was never in the system to begin with.

That is not a claim about our virtue. It is arithmetic. Data you never collected cannot leak, cannot be subpoenaed, and cannot be sold by a future owner of the company. Every one of the incidents above involved an identifier a user handed over at signup.

There is a card path too, because most people want one: £5.49 a month or £54.99 a year, and a 3-day trial that does require card details, because taking cards means taking cards. If you would rather stay unidentified, use the crypto route and skip the trial.

What I am asking you to do

Not to believe me. Read the client. Watch the traffic it generates. Compare the code against what I have written here, and if the two disagree, that is a story and you should tell it.

Every VPN asks for trust at some layer. The only question worth arguing about is how thin you can make that layer. Ours is thinner on the device than on the server, and I am not going to pretend the server half is solved.

See pricing if you want the details, or read the code first. That order is fine by me.