Privacy Policy

Effective Date: August 5, 2026 · Last Updated: October 25, 2026

This policy explains what Privatae LLC ("Privatae," "we," "us") collects, why, and what you can do about it. It covers privatae.ai and the services we operate for you: managed machines, network membership, enterprise services, and Hosts on the Privatae platform.

It is written to be read. If anything here is unclear, email [email protected] and we will explain it.

The Service is offered only to people in the United States. See §11.

1. The short version

• A Host on your own computer sends us nothing about you. It asks us for the bastion's signed release list; anything more goes only if you join it to the network. See §2B.
• On a managed machine, we can reach the server it runs on and that server's console, and the machine sends us a heartbeat about its health, never its contents. See §2A.
• On a Host on the Privatae platform, we can reach into it, because we run the systems it lives on. See §2.
• We do not sell your data. Ever, to anyone, in any form.
• We do not train models on your content. Not ours, not anyone's.
• We collect what running the service requires — who you are, what you paid, how much machine you used, and what our systems log at the boundary.
• You can see it, correct it, export it, and delete it, wherever you live.

2. Hosts on the Privatae platform: what we can reach

This section is about a Host that runs on the Privatae platform, which we offered before October 19, 2026. A managed machine is in §2A, and a Host on your own computer in §2B.

Your subscription gives you an isolated environment. Your files, data and conversations with the assistant live inside it. The code of the applications you build is kept in repositories on a git server we run, unless you move a project's code to a machine of your own; the copy already on our server stays there.

We can reach into it. We run the systems it lives on, and our administrative access to them reaches inside your machine. Our staff and automated checks use that access to run commands inside machines to operate and fix the service. Commands run through the Kubernetes API are logged, but not what is passed to them on standard input, and access that goes straight to a machine's disk is not in that log.

Your machine backs itself up. On the schedule you set (once a day unless you change it), your machine copies its own database and files, encrypts the copy inside itself, and only then sends it where you chose: our object storage, a machine of your own, or storage you rent. Each backup is encrypted to your machine's recovery key, and also to the key of your bastion (the software you run on your own computer) if you have one. You save the recovery key on our account page: your browser makes it there, shows it to you once to save or print, and sends us only its public half. We never receive the recovery key, so we hold no key that opens your backups, and neither does anyone who takes them from us. Your machine uses a recovery key only after you confirm its fingerprint on the machine itself, so we cannot put a key of our own in its place. The page that makes the key is ours, so this rests on that page doing what it says. When a machine changes owner, the previous owner's key stops sealing its backups, and the machine takes none until the new owner saves and confirms their own. If you lose the recovery key and have no bastion key, nobody can open your backups, including you and us. Until you save a recovery key, your machine takes no backups, and it tells you so.

Our older backup job, which we are retiring, works differently, and it still runs for some machines. It uses our access to reach into your machine: every hour it reads your backup schedule, and once a day it copies your machine's database and files onto our backup server, encrypts them there with a key we hold, uploads the encrypted copy to object storage, and deletes the unencrypted working copy when the run ends; if a run is interrupted, the next hourly run deletes what it left. Part of that copy is read straight from disk over SSH. We can open the backups this job makes. We will tell you before it stops for your machine, and the copies it made are deleted when its retirement is complete.

We do not copy the contents of your machine or its backups out for analysis, training, or product work.

Two honest limits, stated plainly because a privacy policy that only lists strengths is not trustworthy:

1. We operate the machine. We can start, stop, migrate, and destroy it.
2. Legal process reaches what we actually hold. We can be compelled to produce account and billing records, and we will comply with valid orders. Because we can reach into your machine, a valid order can reach your machine's contents too. Of your backups, it can reach only the copies made by our older backup job, because those are the only ones we can open. Where we are legally permitted, we will notify you before responding.

2A. C-Series Cloud machines: what we can reach, and what the heartbeat sends

A C-Series Cloud machine is a virtual machine on a server we rent from OVHcloud on our own provider account, which we run for you with C-Series on it. Plainly:

• The server it runs on, and that server's console. They reach the machine's disk, where your Host's data is encrypted under keys only you hold, and can read the machine's memory while it runs unlocked, which holds the key that opens the disk. We use them that way only to recover a machine that cannot recover itself.
• No backup of the machine's disk. We take none. Your Host takes its own encrypted backups once you give it a backup recovery key, and none until then; it keeps them on the machine unless you choose other storage in its settings.
• Your machine's setup code, until the machine reports its setup finished. Then we delete it. Your machines page sends the code to your machine when you press Set up your machine; it is never shown to you.
• The heartbeat. Every five minutes the machine sends us a signed report: its identifier, the time, its software version, whether its Host answers and its deeper health status, free disk space, when its certificate expires, when its last good backup finished and whether backups are paused, how many security updates are pending, whether it needs a restart, when it last installed updates, how long it has been running, whether its disk is locked and since when, whether its setup is finished, and how far it has checked its copy of the access log. It carries nothing from your conversations, memory or files. We keep only the latest heartbeat for each machine, and our servers see the network address it comes from, as they do for any request. When it reports that the machine restarted and is waiting for you to unlock it, or that it was unlocked, that is also written to your machine's access log.
• Updates. The machine installs releases we sign and publish.
• The access log. Our actions on the machine after we hand it to you, the operations our provider recorded on it, and every sign-in to the server it runs on apart from the two accounts our automated programs use, listed on your machines page, newest first.

There is no SSH server and no login for us on the machine, and nothing of ours connects in to it. The passkey and the recovery key that unlock its disk never reach us. Your Host's password and passkeys are kept on the Host. Your model requests go from the machine to the provider you choose, under your key.

2B. A Host on your own computer

C-Series on your own computer, and its installer, ask privatae.ai for the bastion's signed release list, so our servers see that request and the network address it comes from (§3.4). They send us nothing else. If you join it to the network with a membership, the network keeps what it needs to know your Host: its name, the version of C-Series it runs, the certificate it connects with, and when and from which network address it last connected. It also receives what your Host sends to the network's services today: the apps you publish to the App Store for review and what you post to the community board, with your purchases and refund requests in the App Store and your votes and flags on the board. Once they ship, it will also receive the code you push to private git, the sites you host with us, and the backups you store there, encrypted on your Host before they leave it.

3. What we collect

3.1 Account information

Your email address, and a name if you give one. Authentication credentials — passwords are stored only as salted hashes; we never see the original, and where you use a passkey we hold only a public key. Your plan, and your account's status and history with us.

3.2 Billing information

What you paid, when, and for what, and the history of any prepaid balance you held before it was retired.

We do not store your card number. Card details go directly to our payment processor (Stripe). We hold a token, the card brand, and the last four digits so you can tell your cards apart. If you sell through the marketplace, our processor collects the identity and tax information it needs to pay you and to meet its own reporting obligations.

3.3 Operational and usage data

Resource consumption — memory, storage, network volume, machine start and stop times. Metered usage records: which model provider handled a request, how many tokens it consumed, what it cost. The records are counts and costs, not the content of your requests.

3.4 Technical and security data

IP addresses, browser and device information, request timestamps and outcomes, authentication events, and security logs from the network boundary. We use these to run the service, diagnose faults, and detect abuse.

If you ask a question in the Ask box on the wiki (privatae.ai/wiki), with or without an account, we keep the question as you typed it, the answer we showed, the time, the request's IP address and browser details, and the page you came from if your browser says. To write the answer we send your question, with the wiki pages that match it, to a model provider. We use these records to improve the wiki and to limit abuse of the box.

3.5 Network and domain data

Domains you connect or register, DNS records, certificate issuance records, and traffic metadata for sites you publish. If you publish a site, standard web logs for its visitors are generated in the course of serving it.

3.6 Support communications

If you contact us, we keep what you sent so we can help and so we have a record.

3.7 The C-Series Cloud waitlist

If you join the waitlist, we keep your email address, the organization you gave (if you gave one), what you are waiting for, and when you asked. To limit abuse of the form, our server counts requests from each network address (for an IPv6 address, its /64 network) in its memory for at most one hour; the network address is not stored. We use your email address only to send you one email when C-Series Cloud opens. Once it has been sent we delete your address, and in any case we delete it 12 months after you asked.

3.8 Enquiries

If you send an enquiry through the form on the assessment or E-Series page, we keep what you sent, with your name, email and organization, and email it to ourselves so we can reply. It is kept as support correspondence (§10).

3.9 What we do not collect

Apart from the backups described in §2, we do not collect the contents of your machine. We do not run advertising trackers. We do not buy data about you from brokers. We do not build a behavioural profile of you, and we do not perform automated decision-making that produces legal or similarly significant effects.

4. Why we use it

We use the information above to:

• give you the service you paid for and keep your machine running;
• charge you correctly and show you what you spent;
• authenticate you and protect your account;
• detect, investigate, and stop abuse, fraud, and attacks;
• send you service messages — billing, security, renewal reminders, changes to these terms;
• meet legal and tax obligations; and
• understand aggregate load so we can plan capacity.

We do not use it to advertise to you, to sell to third parties, or to train models.

5. Model providers — the part to read carefully

When you use the assistant, your request is routed to a third-party provider of language models, and the content of that request leaves your machine so the provider can answer it. It can include what your machine read to answer you, such as part of a file from a folder you connected.

This is not the only way content leaves your machine:

• your machine's backups leave it, encrypted to your key before they go, and our older backup job copies it while it still runs for your machine (§2);
• the code of the applications you build is kept on a git server we run, unless you move a project's code to a machine of your own (§2);
• posts to a circle are stored by the platform and shown to that circle's members;
• a message from your machine to another machine on the service goes through the platform, which carries its text, when your machine holds no permission slip for it or the direct connection fails; and
• if you set your machine to store its context in your own S3 bucket or on a computer of yours that has joined your network, that content is kept there.

What this means:

• The content of your request goes to that provider to be processed.
• That provider's own terms and privacy practices apply to what they receive. They are not ours and we do not control them.
• We pass through what the request requires. We do not attach your name, email, or billing identity to it.
• You can bring your own provider account. If you connect your own key, your requests go to that provider under your own agreement with them, and we are only routing.
• You can use models running on machines you control, in which case the content does not go to a third party at all.

Current providers are listed on the platform and may change. If you want a request never to reach a third party, use a local model or do not make the request.

6. Who else touches your data

We use a small number of service providers to run the business. They get only what their job requires, are bound by contract, and may not use it for their own purposes.

Beyond these, we share only: when you tell us to; when a valid legal order requires it (§2); to protect against imminent harm; and to a successor if the business is sold, in which case we will tell you first and this policy continues to apply until replaced on notice.

We have never sold personal information and we do not intend to.

7. How it is protected

Data is encrypted in transit. Machines are isolated from one another, and the boundary between them is enforced by the platform, not by convention. Network egress is default-deny. Administrative access to infrastructure is restricted, authenticated, and logged. It includes a path into your machine, which our staff, our automated checks and our older backup job use (see §2). A command run inside your machine through the Kubernetes API is logged, but what is passed to it on standard input is not, and the parts of a backup read straight from disk over SSH do not pass through that log.

Backups exist so that a failure does not lose your work. Your machine's own backups are encrypted inside it before they leave, and those it keeps in our storage expire on a schedule. Our older backup job encrypts its copies on our backup server before they are uploaded; the working copy it makes on that server is written without that encryption, and is deleted when the run ends.

No system is perfectly secure, and we will not tell you otherwise. If a breach affects your personal information we will notify you and the relevant authorities as the law requires, without undue delay.

8. Where your data lives

Infrastructure is located in the United States. We do not replicate customer data to servers outside the US. A managed machine runs in the region you pick from those we offer: Virginia for a virtual machine, and Virginia or Oregon for a bare-metal server, which is ordered by email.

Third-party model providers may process requests on their own infrastructure, which may be located elsewhere — see §5.

9. Cookies

We use cookies that are necessary to make the site work: keeping you signed in, maintaining your session, and security protections such as CSRF tokens.

We do not use advertising cookies, third-party trackers, or cross-site analytics. There is no ad network on this site and no data broker receiving a feed from it.

Because we set no non-essential cookies, there is no consent banner. Blocking essential cookies will stop you being able to sign in.

10. How long we keep things

After the 90-day window, container data is permanently and irrecoverably deleted. We cannot restore it, and neither can you. Export before then.

11. Availability — United States only

We offer the Service only to individuals and entities in the United States. We do not offer it in the European Union, the United Kingdom, or elsewhere, and this policy is written for US law. If you are outside the US, do not use the Service.

Because we do not serve the EU or UK, this policy contains no GDPR section, no EU representative, and no international transfer mechanism. We would rather say that plainly than publish a compliance posture we do not maintain.

12. Your rights

Twenty states now have comprehensive privacy laws, and their rights differ in detail. Rather than make you work out which applies to you, we give every customer the same rights, wherever you live:

• Know what we hold about you and why.
• Access a copy of it.
• Correct anything inaccurate.
• Delete it, subject to records we must keep by law (§10).
• Port it — get it in a portable, machine-readable format.
• Opt out of any sale or sharing of personal information for targeted advertising, and of profiling with significant effects. We do none of these things, so there is nothing to opt out of — but the right stands.
• Not be retaliated against for exercising any of the above. We will not degrade your service or change your price because you asked.

12.1 How to exercise them

Email [email protected]. We will verify you through your account email and respond within 45 days, extending once by a further 45 where a request is complex, and telling you if we do.

Much of this you can do yourself, immediately, from your account — export and deletion are controls, not requests.

An authorised agent may act for you with written permission we can verify.

12.2 Appeals

If we refuse a request, we will tell you why, and you may appeal by replying to that decision. We will respond to an appeal within 45 days. If we refuse again, we will tell you how to complain to your state Attorney General.

12.3 California

For California residents, the categories of personal information we collect are in §3, our purposes in §4, and recipients in §6. We have not sold or shared personal information in the preceding twelve months, and we do not collect personal information from anyone we know to be under 16. Sensitive personal information is used only to provide the service and never to infer characteristics about you.

13. Children

The Service is for adults 18 and over. We do not knowingly collect personal information from anyone under 18. If we learn that we have, we will delete it and close the account. If you believe a child has given us information, email [email protected].

14. Your own visitors

If you publish a website or an application, you are its operator and the data you collect from your visitors is yours to govern. You need your own privacy policy and your own legal basis for what you collect. This policy covers our relationship with you, not your relationship with the people who use what you build.

Published sites are not search-indexed by default; turning that on is your decision, per site.

15. Changes

If we change this policy materially we will email you and post the change at least 30 days beforehand, unless a change must take effect sooner for legal reasons. The "Last Updated" date always reflects the current version, and we keep prior versions available.

16. Contact

Privatae LLC — Wyoming, USA

• Privacy: [email protected]
• Security: [email protected]
• Legal: [email protected]

If you are unsatisfied with our response, you may complain to your state Attorney General.