Skip to main content

Privacy policy

Version 2026-11-20, last updated 2026-11-20

This policy explains which personal data zdr.sh processes, why, on what basis, for how long and with whom, and which rights you have. It applies to the website, the dashboard, the API and support. It is published in English, which is the only version.

1. Who is responsible#

zdr.sh is operated by the operator of zdr.sh. Write to support@rackduck.com about anything in this policy, including to exercise your rights. The operator is the controller of the account, billing and security data of every customer, and of any content a consumer chooses to store. For content a business customer sends through the API, the operator acts as the processor of that customer; a business customer can ask at the same address for the data processing agreement we offer under Article 28 of the GDPR.

2. What data we process#

  • Account and security data: name, email address, the password kept only as a hash and never in clear, the secret of the one-time codes of your authenticator app if you enable it, the public keys of your passkeys, the language, the security settings, the active sessions with their IP address and browser, and the history of your consents with date, time and IP address.
  • Billing data: the amounts, assets, networks, transaction ids and statuses of your payments, the address a refund is paid to, and the billing profile an earlier version of the service collected (a customer type, a name or company name, an address and any tax identifiers given), kept as the record of the payments it was collected for. No card data and no bank details: we take no card and no bank transfer, so none exists to store.
  • Payment data: the address the top-up screen shows you and, for Monero, the payment id that goes with it; the address you name for a refund; the asset and network of each transfer, and the confirmations it reached. A crypto address is pseudonymous, not anonymous: it is personal data when it can be linked to you, and we treat it as such.
  • Usage data: for each API request, the account, the key, the model, the tokens, the latency, the cost and the outcome, plus the attribution the client sent with it (an end-user id and the entries of its metadata, bounded and never the request body), written to our analytics store. The usage records never contain the text of the prompts.
  • Your content: the text, files and images of the requests and responses, processed while the request runs. We keep no copy of it except in the memory vault you write to, the response cache if you turn it on, and the idempotency replay of a request that carries an Idempotency-Key.
  • Server logs: the IP address and, when you are signed in or use a key, the account id of each request to the website, the dashboard and the API, with its time and outcome. Logs never contain request bodies.
  • Support data: what you write to us and our answers.

We estimate the country of an IP address with a database installed on our own servers, without sending the address to anyone: this product includes IP geolocation data created by DB-IP (https://db-ip.com), licensed under CC BY 4.0.

3. Why we process it, and on what basis#

  • Contract: to open the account, bill the usage, serve the requests and answer support. The account and the billing data are necessary to make the contract: without them we cannot provide the service.
  • Legal obligation: the accounting and tax rules, which require keeping the records of the payments for the period the law sets.
  • Legitimate interest: to keep the service reliable and secure, through the server logs and the usage records of section 2. You have the right to object at any time.
  • Consent: for anything marked optional, such as product emails. You can withdraw a consent at any time, with effect for the future.
  • We do not process special categories of personal data on purpose: the acceptable use policy prohibits sending them. No automated decision with legal or similarly significant effects is taken about you (Article 22).

4. Zero retention, and where it stops#

  • We do not collect trace copies of requests and responses. There is no trace store, no sample of traffic kept for quality, no automated scrubbing pipeline and no opt-out to look for, because nothing is collected in the first place.
  • The text of a request and the text of an answer exist while the request runs and nowhere on our side afterwards, except in three places that section 2 names: the memory vault, the response cache and an idempotency replay. Each of the three is a choice, not a default: the vault holds only what you write to it, the response cache is off unless a key turns it on and holds an answer for ten minutes, and an idempotency replay holds the answer to a request that carried an Idempotency-Key for 24 hours so a retry is answered once. None of the three is read to train, score or review a model, and none is shared with anyone.
  • Serving happens on two tiers. On a zero-retention route, a partner provider serves the request as our processor under written data protection terms, including a commitment not to keep the request or the answer after serving it. A narrow security or abuse inspection carve-out reserved by the partner's terms may apply; ask us for the terms in force.
  • On a transient route, served on rented third-party GPUs, the platform persists nothing, but the third-party operator of the machine may access memory while the request is processed. The catalog labels these routes as transient, and a key restricted to zero-retention routes never reaches one.
  • Not every route is zero retention, and we do not claim that it is. The label on the route and the response tells you which tier served the request.
  • Because no request content is kept, there is nothing to erase from a trace store, and no dataset can be built from your traffic.

5. No datasets#

We do not build, sell or license a dataset of requests or responses, and no recipient receives one. The dataset recipient licence that an earlier version of this policy mentioned is withdrawn: it covered traces that are no longer collected. If that ever changed, this policy would name the recipients one by one and ask for a separate consent before any data reached them, and nothing in these documents would be read as a consent given in advance.

6. Who receives the data#

RecipientWhat for
upstream model providers and the operators of rented GPUs, in the countries where they processserving the requests and generating the responses
hosting and object storage providerservice infrastructure and encrypted backups
SMTP providersending email
web search providerrunning the queries of the search tool
accounting providerbookkeeping of the payments the law requires us to keep
authorities, when the law requires ittheir requests
the account's usage webhook endpointonly when the account configures it, to the address the account names

A business customer can ask at support@rackduck.com for the data processing agreement we offer under Article 28 GDPR: it names the sub-processors and the transfer safeguards of each category, and it is concluded in writing outside the platform.

Transfers are handled by category of recipient: a recipient that processes inside the European Union needs no transfer tool, and for a recipient that processes outside it the transfer relies on an adequacy decision of the European Commission where one exists, and otherwise on its standard contractual clauses with the supplementary safeguards needed. Ask us for a copy of the tools in force per category. The current sub-processor list, with the providers active for your requests and the countries where they process the data, is available on request from support@rackduck.com.

7. How long we keep it#

DataRetention
Account and security datafor the life of the account; after closing, what proves the contract stays for the accounting records and the rest is deleted within 90 days
Sessions (IP address and browser)until the session expires, at most 30 days after last use
Billing, payments, refunds, withdrawals, consent history10 years
Per-request usage dataup to 400 days
Memory vaultuntil you delete the notes or close the account; a note you delete, and its versions, are removed 30 days after the deletion
Response cache, if turned on10 minutes
Idempotency replay24 hours from the settlement
Usage webhook delivery records30 days from the delivery
Support messagesup to 24 months from the end of the conversation
Backupspersonal data erased in the service stays in our encrypted backups for up to 5 weeks, until those backups expire
Server logs (IP address, account id)up to 30 days in our log store, plus a rolling buffer of at most 250 MB per service on the server that newer lines overwrite

8. Your rights#

You can access your data, rectify it, erase it, restrict its processing, receive it in a portable format, and object to processing based on our legitimate interest. Where something rests on consent, you can withdraw it at any time. You can close the account from the dashboard or by writing to support@rackduck.com. Exercise your rights in the dashboard where a control exists, or write to support@rackduck.com: we answer within one month, and where the dashboard has no control we provide the data or the copy by hand. You can also complain to the supervisory authority for data protection of the country where you live or work.

9. How we protect it#

We protect data by encrypting the connections in transit and the backups at rest, by storing passwords only as a hash, by sealing secrets with the server key, by restricting access to authorized staff, and by keeping request bodies out of the logs. The service's disks are not encrypted at rest: what protects the data stored there is access control on the host and on the database. In case of a personal data breach we notify the competent supervisory authority within 72 hours where the law requires it, and we tell you without undue delay when the risk to your rights is high.

10. Cookies#

The website and the dashboard use two technical cookies and a few preferences in your browser, and nothing else: the cookie policy names them one by one. No analytics, no advertising, no third-party tracking.

11. Contact and changes#

The version and date at the top identify the text in force; material changes are announced by email before they apply, and no new purpose or category of data is added without a new consent where consent is the basis. For questions about this policy or your data, write to support@rackduck.com.

This site uses only the technical cookies the sign-in and the security need. No analytics, no advertising, no tracking. Cookie policy