How durin handles personal data.
Last updated November 5, 2026. This notice explains what personal data Marathon Digital Pte Ltd. collects through the durin website, application, support channels, and governed gateway, how long we keep it, and how to exercise your rights. Where we process data for a customer, that customer’s signed agreement or data processing addendum also applies.
Who is responsible for your data
durin is operated by Marathon Digital Pte Ltd., a Singapore company (UEN 202023814G). For purposes we decide, such as running accounts, billing, support, sales, and this website, Marathon Digital is the controller under the EU and UK GDPR where they apply and the organisation responsible under Singapore’s Personal Data Protection Act (PDPA). John Rocela, Founder, runs the business and is accountable for security, privacy, and supplier decisions. Contact our data protection officer at privacy@getdurin.com, or email contact@getdurin.com about anything else.
Compliance commitments
We are committed to meeting applicable GDPR and Singapore PDPA obligations and aligning our security programme with ISO/IEC27001 principles. These are management commitments, not certification or independent assurance that every obligation has been fulfilled.
Organization usage analytics
We collect daily organization totals to operate durin, understand service usage and monitor business operations. These include governed request outcomes, approval totals, counts of members, clients and connections, and the organization’s package and billing status. Cloudflare Basin stores these totals under a random organization key, without organization names, person identifiers, request arguments, tool results or conversation content. Counts of people below three are withheld from the lake. This operational collection is separate from optional PostHog analytics. Facts are kept for up to 25 calendar months and can be deleted earlier when an organization is erased. Collection stops during the deletion process; older snapshots and unreferenced files remain until Cloudflare maintenance removes them, and provider cleanup stays open until removal is verified.
Combined analytics reports
We also store combined daily analytics totals in Google BigQuery in Singapore. Each published group includes at least five organizations and contains no organization keys, person identifiers or request content. We rebuild these reports from eligible retained facts so organization erasure and the 25-calendar-month limit apply to current reports. Google recovery copies can remain after a report is replaced or deleted.
How we protect personal data
Access to workspace data requires current identity, membership, and role checks. Selected connection and export-destination credentials are encrypted with AES-256-GCM, and governed requests leave append-only audit records. These measures are not independently audited: no SOC 2 attestation, ISO certification, or other independent assurance is claimed.
Website and application analytics
PostHog product analytics is optional and stays off unless you choose it in the consent banner or in Privacy settings. With analytics on, PostHog receives page views, link clicks, MCP simulator completions and successful lead submissions from the website. Lead events include the form type and a random submission identifier, never contact details or form answers. Audit pages send only successful report-request conversions; they do not send audit answers, configs, facts, grades, reports or verification tokens, and verification pages do not initialize analytics. Successful sales inquiries in the application use the same lead event. From the application it receives page views and page leaves, the onboarding and connection setup steps you view, select, cancel or abandon (with the provider involved), and the method, route and status of requests other than GET. URL query strings and fragments are removed. With analytics on, PostHog also stores an identifier in a cookie shared across getdurin.com subdomains and in browser storage; in the application, events can be associated with your user and organisation, and signing in links earlier visits from the same browser to your user in PostHog. Server-side events (sign-up, invitation, onboarding, billing and connection milestones, and the method, route and status of requests other than GET) are sent only for people who allowed analytics under the current notice version, and only for events after that choice. Our servers keep the analytics choice for your account, not for one browser. The application sends your browser’s stored analytics decision with each request you make while signed in, and for server-side events the most recent decision applies. A browser with no stored choice and no privacy signal sends nothing, so server-side events can continue for requests it makes while your account holds a grant from an earlier decision. A choice you make on the website is adopted for your account when you sign in to the application. Session recording and replay, and PostHog’s automatic capture of clicks and other interactions, are not used. We use Sentry for error diagnostics, which run whatever you choose. Error reports from our servers for onboarding, billing and connection steps carry your user ID. With analytics on, the application in your browser also gives Sentry your user ID and adds breadcrumbs for page views, setup steps and requests other than GET to its error reports; turning analytics off or signing out stops this in that browser. Diagnostic filtering reduces sensitive fields but does not guarantee that every error message is free of personal data. Outside audit pages, analytics also measures scroll depth and page loading, responsiveness and layout stability. These measurements exclude page text, element details and resource URLs.
Campaign attribution
Campaign attribution is optional and stays off unless you choose it. With attribution on, a first-party cookie named durin_acquisition keeps the campaign tags of the link you followed, the referring site’s host name and the landing page path for 30 days. When you sign up, the campaign record (if attribution is on) is stored with the new account so the sign-up can be credited to the campaign, and the browser’s analytics identifier (if analytics is on) is stored with it. If analytics is also on, the sign-up events sent to PostHog include that campaign record, and successful lead submissions include the campaign source, medium and name. With attribution on when you submit a lead form, we also save the first visit and most recent non-direct visit, including campaign tags, referring host, landing path, visit time and submission page, with the lead and email these details to our team. This applies to sales, design partner, SAFR, ISO42001, demo, risk audit and partner application forms. These details follow the lead’s retention rules; team mailbox copies require separate cleanup. Turning attribution off stops future capture and clears the browser cookie; it does not erase details already submitted. With analytics also on, website page views send PostHog the campaign source, medium and name, and the referring site’s host name. These page views exclude campaign search terms and ad click identifiers.
Google Analytics
If you allow Analytics under the current notice, marketing and documentation pages load Google Analytics through Google Tag Manager. Google may receive cookie identifiers, page visits, and browser and network information. This integration does not send lead form fields, audit content or account identifiers. Audit, report and verification pages do not load this container. Turning Analytics off clears first-party _ga, _gid and _gat cookies and browser storage accessible to this site, and reloads the page to stop the tag. This does not erase information Google already received; Google’s retention settings and privacy rules apply.
Google Ads measurement
If you allow Campaign attribution under the current notice, public marketing pages load Google tag AW-18540122999. Google may receive advertising click and cookie identifiers, page visits, and standard browser and network information to measure our advertising. We disable personalized advertising and enhanced conversions and do not pass form fields, contact details, audit content or account identifiers to this tag. Audit, report and verification pages do not load it. Turning Campaign attribution off denies Google advertising consent, clears first-party _gcl_ cookies and storage available to this site, and reloads the page to stop the tag. This does not erase data Google already received; Google’s own retention and privacy rules apply. With Campaign attribution on, an accepted demo request also sends a conversion event and a random submission identifier to Google to avoid duplicate counts. This records the form submission, not a confirmed calendar booking.
Sales inquiries and design partner applications
Sales, design partner, SAFR and ISO42001 inquiries save your name, email, optional phone number, company name, company size, message and submission source in our database for 90 days. We email these details to our team using Cloudflare Email Service. Form submissions are not sent to Sentry. Cloudflare Turnstile verifies these forms. We use a keyed hash of your IP address to limit submissions to one every five minutes; the temporary receipt contains no form fields or raw IP address. Correspondence is separate from error diagnostics. Do not include passwords, access tokens, or sensitive customer content in your message.
ISO42001 tracker downloads
To download the ISO42001 implementation tracker, we ask for your name, work email, optional phone, company, company size and privacy consent. We save these details, consent version, submission time and download source in our database for 90 days, and notify our team by email. Optional tracker details, such as a programme owner, plan and audit dates and AI roles, and your checklist answers are used only to prepare the file: we do not save them or send them to analytics. We do not subscribe you to marketing through this form.
Demo requests
We save demo requests in our lead database on Cloudflare and email them to our team, including your work email, organization name, role, organization size, and how you heard about durin. Form submissions are not sent to Sentry. Cloudflare Turnstile verifies the form. After we accept your details, Google Calendar loads so you can choose and confirm a time. Google receives your browser’s IP address and any booking information you enter in its calendar.
Demo account setup
A demo request also starts account signup with your work email and organization name. We send a verification email before showing the calendar. Your account is created after you verify your email and set a password. Existing accounts are kept as they are. If a referral is applied, it follows this signup. The signup verification retention schedule applies to pending requests and verification emails.
Partner applications and referrals
Partner applications send your name, work email, company, country, partner type, customer segment, expected volume, optional website and phone, description, the terms and privacy versions you acknowledged, and your marketing choice to our control database for review and email them to our team; Cloudflare Turnstile verifies the submission and a keyed hash of your IP address limits submissions. Rejected applications have their contact details removed 365 days after the decision. A referral link sets a first-party cookie on this website, __Host-durin_referral, that holds a signed referral token and lasts up to 90 days. It is set when you follow the link, whatever you chose in Privacy settings, and the banner shown while it applies lets you remove it. If you follow a sign-up link while it applies, the referral is handed to the application, which stores a record of it and sets a cookie of the same name on its own host. A referrer only sees a masked organization alias, dates, referral status and their own commission amounts. Referral data is not sent to analytics.
Privacy settings and browser signals
You can change your analytics and attribution choices at any time in Privacy settings: in the footer of this website and, in the application, in the account menu, in personal settings and in the privacy links on pages such as sign-up, organisation registration, onboarding and earnings. Your choice is stored in a first-party cookie named durin_consent for 180 days, shared by getdurin.com and app.getdurin.com; cookies that keep the service working, keep you signed in and remember this choice are always on, as is the referral cookie described under Partner applications and referrals. Turning analytics off stops further collection, clears analytics storage and reloads the page if analytics was running on it; turning attribution off removes the durin_acquisition cookie. A Global Privacy Control or Do Not Track signal from your browser keeps both purposes off. It does not affect Sentry diagnostics or support chat that you open.
Support chat
Support chat is provided by Crisp. Its chat bubble loads automatically on the website and application and can send browser details and IP addresses to Crisp. Messages you send are processed by Crisp; in the application, your account email is shared when you open chat so support can identify you. Crisp can use cookies and browser storage. In the application, Stop support chat in Privacy settings reloads the page, stops loading Crisp in that browser tab, and removes its cookies and stored data from the browser. It also stops chat in other open tabs of the application. Signing out of the application resets the chat session. You can email contact@getdurin.com instead. Do not submit passwords, access tokens, or sensitive customer content in chat.
Account, billing, and service providers
We process identity and membership information to manage access, account and subscription information for billing, and correspondence to respond to enquiries. The service integrates Cloudflare for hosting, storage, form verification, and email delivery, including design partner applications; WorkOS for identity; Stripe for billing; Sentry for diagnostics; PostHog for optional analytics; and Crisp for support chat. Browsers also load a typeface from Google Fonts, which receives the browser’s IP address. If an administrator turns on AI policy drafting, the policy description they write and the workspace’s connection, tool, and client identifiers are sent through OpenRouter to a model provider that is required to keep no copy, and durin does not store the description or the response. Customer-selected upstream systems and export destinations have their own processing boundaries. This inventory is not an executed subprocessor schedule. The providers that process customer personal data under a data processing addendum are listed in the subprocessor schedule at /privacy/subprocessors.
Governed gateway data
For each governed request, durin records who made it, the client and connection used, the allow, deny, or approval decision, and the execution outcome. This metadata can itself be personal or confidential data. Tool requests and results pass through durin to complete the request and are not kept afterwards unless an administrator turns on conversation capture. If an administrator selects Durin Smart Evaluation policy evaluation, tool requests and results are also sent to Durin Smart Evaluation for evaluation before they are dispatched or returned. Separately, when the evaluation service is configured, allowed or approval-required write actions send the tool name, upstream name, description, connection provider and exact arguments for a background reversibility report in either evaluation mode. This reporting is always enabled and does not change access, approvals or execution. durin does not host or train AI models; results returned to an external client can reach that client’s model provider.
Optional conversation capture
Conversation capture defaults to metadata only: submitted message content is discarded before persistence. When an administrator enables redacted message capture, new captures are retained for 7, 90, or 180 days, or 360 days on Enterprise. Workspaces with an older 1–30-day setting keep it until an administrator selects a supported period. Expired messages become unreadable. Scheduled cleanup removes retained payload objects, except those covered by a retention hold, which stay until the hold is released and remain unreadable after their expiry. These periods do not cover audit metadata, account records, billing, backups, analytics, or support messages. Redaction is pattern-based and is not complete data-loss prevention.
Processing roles and international transfers
For customer-directed processing, we would ordinarily act as a GDPR processor or Singapore PDPA data intermediary; for purposes we determine, we may act as controller or organisation. The actual processing and agreement determine the role. Our service providers process personal data in other countries, including the United States. No EU-only or Singapore-only residency guarantee is offered. For customer content, the data processing addendum at /privacy/dpa sets out our role, customer instructions, subprocessors, international transfer safeguards, breach notice, and deletion or return; it takes effect when both parties sign it. The subprocessor schedule at /privacy/subprocessors lists each subprocessor, where it processes data, and the safeguards that apply. Contact privacy@getdurin.com to discuss legal holds, backup expiry, tenant deletion, and international-transfer requirements.
How long we keep personal data
We keep personal data only as long as we need it for the purposes in this notice, or longer where the law requires it. The schedule below shows how long each category is kept and what happens next. Data we hold for a customer also follows that customer’s instructions and agreement. A legal claim, regulatory request, or security investigation can require us to keep specific records longer; we keep them only for that purpose.
- Organization usage analytics
Up to 25 calendar months
Daily organization totals in Cloudflare Basin use a random organization key and contain no person identifiers or request content. Organization erasure can delete facts earlier. Older snapshots become eligible for expiration after seven days; physical file cleanup is verified separately. The organization-to-key mapping remains as erasure evidence so delayed deliveries can be removed again.
- Combined analytics reports
Up to 25 calendar months
We also store combined daily analytics totals in Google BigQuery in Singapore. Each published group includes at least five organizations and contains no organization keys, person identifiers or request content. We rebuild these reports from eligible retained facts so organization erasure and the 25-calendar-month limit apply to current reports. Google recovery copies can remain after a report is replaced or deleted.
- Account and workspace records
Until the workspace is deleted
Names, email addresses, memberships, roles, and workspace settings. When a workspace or a person’s account is deleted through our deletion process, access stops immediately and these details are scrubbed from workspace records after a 7-day grace period. Internal identifiers remain so audit and billing records stay consistent.
- Signup verification
Verification links expire after 3 days
When you finish signing up, or the link expires unused, we clear the email address and verification token from the signup record. The workspace name you entered and a one-way hash of the address, which prevents duplicate signups, stay on the record. Records of the verification emails we send are deleted after 30 days.
- Audit records
No routine expiry
Who made each governed request, the client and connection used, the allow, deny, or approval decision, the outcome, and a digest of the arguments rather than the arguments themselves. Audit records are append-only so customers can rely on them as evidence, and they remain after a workspace is deleted. Audit export files expire within an hour.
- Approval memory
No routine expiry
Approval memory is on unless an administrator turns it off. Each reviewer’s decision is remembered with a one-way sketch of the request: argument names, value types, and salted hashes, never argument text. Reviewers see the closest earlier decision on a similar request. An “Allow always” decision can approve closely matching requests for 30 days by default, and where an organization opts in, a remembered denial can deny an exact repeat for 7 days. A policy change, a requester or reviewer losing membership, turning memory off, or an administrator’s revocation ends a remembered decision, and the entries remain as evidence with the audit records.
- Tool requests and results
Not kept after the request completes
Content passes through durin to the connected system and back to your client. It is not stored unless an administrator turns on conversation capture.
- Conversation capture
7, 90, or 180 days, or 360 days on Enterprise
Capture keeps metadata only unless an administrator turns on redacted message capture for one of these periods. Captured messages become unreadable when their period ends, and scheduled cleanup deletes the stored content; content under a retention hold stays, still unreadable, until the hold is released. Workspaces with an older 1–30-day setting keep it until an administrator selects a supported period. Capture metadata, such as names and timing, stays with the workspace’s records.
- Notification emails
Recipient address cleared after 30 days
For workspace notification emails, we clear the recipient address and the email provider’s message reference 30 days after the email is queued, and delete delivery records after 90 days.
- Partner applications and referrals
Rejected applications: 365 days after the decision
We then remove the applicant’s contact details. Accepted applications stay with the partner account until it is deleted. Unclaimed partner invitations expire after 14 days, and we then remove the invited email address. A referral link’s cookie expires after 90 days.
- Sales inquiries, demo requests and design partner applications
90 days after our last exchange
We keep sales, demo, design partner, SAFR and ISO42001 database records for 90 days after submission. We email lead notifications to our team. Email correspondence is kept for 90 days after our last exchange. If you become a customer or design partner, we keep the correspondence your account needs. Each form’s rate-limit receipt expires after 5 minutes.
- Support chat
180 days after the conversation closes
Messages you send and the browser details Crisp collects. In the application, Crisp also receives your account email address.
- Product analytics
Up to 30 days
PostHog events from the website and application while analytics is on, and our record of the product milestones we send to PostHog. The campaign cookie is set only while attribution is on and expires after 30 days. Global Privacy Control and Do Not Track keep both off.
- Error diagnostics
Up to 90 days
Sentry error reports from the website, application, and services, filtered to reduce personal data.
- Billing and payout records
As long as Singapore law requires, at least 5 years
Subscription, invoice, payment, commission, and payout records kept for tax and accounting. Stripe and our payout provider hold payment details under their own terms. We delete our copies when the legal retention period ends.
- Backups
Up to 30 days
Cloudflare keeps point-in-time recovery copies of our databases. Deleted data remains in these copies until they roll off. If we restore from a copy, we reapply deletions before the service resumes.
Your privacy rights
You can ask for a copy of the personal data we hold about you and how we have used or disclosed it, and ask us to correct data that is inaccurate or incomplete. Where we rely on your consent, you can withdraw it at any time by telling us; we will tell you about any likely consequences and stop the processing that relies on it. If the EU or UK GDPR applies to our processing of your data, you can also ask us to erase it, restrict how we use it, or provide it in a portable format, and you can object to processing based on our legitimate interests or for direct marketing. These rights have legal limits; if one prevents us from meeting a request in full, we will explain why.
Children
durin is directed at business users. We do not knowingly collect personal data from children under 16, or a higher age where local law requires it. Contact privacy@getdurin.com if you believe we have collected such data.
Marketing and Do Not Call
We use sales inquiry details to respond to your request and, where permitted, to follow up about durin. You may opt out of marketing emails by emailing privacy@getdurin.com. If we call Singapore telephone numbers for marketing, we will comply with the PDPA Do Not Call provisions.
How to make a request
Email privacy@getdurin.com, from the address linked to your account where possible, and tell us what you are asking for. We may ask for information to confirm your identity, or a representative’s authority to act for you, before we act. We respond without undue delay. Under the GDPR, we respond within one month, which we can extend by two further months for complex or numerous requests if we tell you why within the first month. Under the PDPA, if we cannot respond to an access request within 30 days, we tell you in writing within that time when we will. Do not email secrets, tokens, regulated payloads, or customer content.
Requests about customer content
If your personal data is in a workspace that a customer controls, such as your employer’s, that customer decides how it is used. Send your request to the customer. If you contact us instead, we will refer you to the customer where we can and help the customer respond, as our agreement with them and the law require.
Personal data breaches
If we have reason to believe a breach has affected personal data we control, we assess it promptly. Where the PDPA requires it, we notify the Personal Data Protection Commission (PDPC) and affected individuals; where the EU or UK GDPR requires it, we notify the relevant supervisory authority and affected individuals. Each notification is made within the time the applicable law sets. If a breach affects personal data we process for a customer, we notify that customer without undue delay after we become aware of it or have reason to believe it occurred, so the customer can meet its own obligations. Notices describe what happened, the data involved, the likely consequences, and the steps we and you can take. Report a suspected incident to privacy@getdurin.com.
Complaints
We would like the chance to resolve any concern first. You can also complain to Singapore’s Personal Data Protection Commission or, where the EU or UK GDPR applies, to the data protection supervisory authority where you live or work or where you believe an infringement occurred.
Changes to this notice
We update this notice when our processing changes, and the date at the top of this notice shows when it last changed.
Contact us
Marathon Digital Pte Ltd., a Singapore company (UEN 202023814G), operates durin.