Legal

Dokima Privacy Notice

Last updated: 11 November 2026

1. Who we are

Northloom Ltd (company number SC895860) is the controller of the personal data described in this notice. Northloom Ltd is registered in Scotland at 5 South Charlotte Street, Edinburgh, EH2 4AN. In this notice, “Northloom”, “we”, “us” and “our” refer to Northloom Ltd. Dokima is a Northloom product.

Contact us about privacy or legal matters at contact@northloom.ai. Product support is available at support@dokima.net and security reports may be sent to security@northloom.ai.

2. Scope

This notice covers personal data we process in connection with dokima.net, sales and trial applications, customer administration, licensing, billing, support, product-security reports and business communications.

Dokima is a business-to-business product. It is not intended for consumers or children.

3. Information we collect

Depending on how you interact with us, we may collect:

  • Business contact data: name, work email address, telephone number, job title, employer, business address and correspondence.
  • Sales and trial data: organisation details, intended use, repositories and team size, procurement information, application answers and our assessment or approval notes.
  • Account and licence data: username, stable internal account identifier, password hash, optional passkey public keys and credential metadata, email-verification records, invitations, memberships, roles, private publishing installation identifiers and access/audit events; customer organisation, plan, authorised contacts, licence or entitlement identifiers, activation and expiry information, product version and basic service events needed to provide and protect the licence.
  • Billing data: billing contact, billing address, tax information, subscription, invoice and payment status. Stripe processes full payment-card details; we do not receive or store the full card number or security code.
  • Due-diligence data: information required for customer, sanctions, restricted-party, destination and end-use screening, including information about organisations, directors, beneficial owners or authorised representatives obtained from you or reputable public and commercial sources.
  • Support and security data: support requests, diagnostic material you choose to provide, communications, reports submitted to us about the security of Dokima or our services, and our response records.
  • Website and security data: IP address, date and time, requested page, user-agent and related server or security logs. The website records a small set of first-party funnel paths in those ordinary server logs (for example, that a visitor opened the trial flow or advanced a form step). Fixed aggregate page-view and CTA labels record allowlisted incoming sources (such as Google, Reddit, LinkedIn and Northloom referrals), without recording arbitrary URLs, query values or visitor identifiers. Fixed promotion labels also record Krasis download-page promotion arrivals, without storing campaign query values or linking them to an account. These event requests suppress the referring page, contain no form contents, email address, licence credential or Checkout token, and use no analytics cookie or external tracking provider. The website currently requests font files from Google Fonts, which may receive request information such as an IP address and browser details.
  • Signup and product lifecycle data: first-party signup stages and fixed failure categories, linked through a random page-lifetime journey ID and internal account/organisation IDs. Updated Dokima clients also send setup completion, review-job starts, accepted completions, coarse failure categories and successful report generation, authenticated to the licensed installation. Product events may also include anonymous finding statistics: counts by severity and validation status, with keyed one-way identifiers that let us count the same codebase or finding only once across reruns and installations. Product events contain no source code, finding titles or descriptions, paths, prompts, model output, credentials or form contents. We may publish combined finding totals across all installations, such as on dokima.net, but never figures that identify a customer, organisation or codebase. They operate independently of private report publishing. Set DOKIMA_TELEMETRY=off in the environment running Dokima, including its background worker, to disable product events. No analytics cookies or external analytics provider are used. Events are retained for up to 90 days, capped at 200,000 events per backend, to understand onboarding problems and improve reliability.
  • Marketing preferences: whether you wish to receive or stop receiving relevant product communications.

Updated clients queue these limited product events locally for up to seven days (at most 2,048 events) and deliver them in compressed scheduled batches. Delivery receipts prevent duplicate records on retry and are retained for up to eight days. Private report sharing remains a separate explicit choice; disabling telemetry excludes events from exchanges, and disabling sharing excludes reports and review status. The upload interval is at least five minutes and may be increased centrally up to four hours. Older clients retain their existing delivery behaviour until updated.

4. Source code, findings and AI runners

Dokima is installed and run in the customer’s environment. Local HTML reports are always generated. Before any optional private-team upload, a user confirms either private Dokima.net publishing or local HTML reports only. Existing installations without a choice receive the same prompt. Selecting an option by default does not itself start an upload.

When private publishing is enabled and linked to an organisation, Northloom receives worker status, the latest vulnerability report and Keyholder Report, evidence/source excerpts, repository/component and code-location references, and questions, answers and discussion history between authorised team members and the linked installation. Question and answer attribution includes the team member’s account identity; local contributions are identified as installation-user statements. Repository checkouts, runner prompts, credentials, raw runner logs and complete local scan state are not uploaded. Local-only operation keeps reports and discussions on the customer’s machine. Customers should minimise personal data and secrets in questions, answers and evidence.

All authorised organisation members may ask and answer questions for reports they can access. Installation publishing credentials can receive only discussion for their own publication and delete that publication; they cannot read reports, account, billing or membership data. The customer’s chosen AI runner may receive relevant source and attributed discussion as review context under the provider’s independent terms and privacy practices. An opaque random workspace identifier and content-free daily usage counter enforce the shared question limit across resets and re-pairing. Old UTC-day counters are pruned separately from discussion content.

Hosted reports are available only to invited organisation members and agents using API keys created by an organisation owner or admin. Organisation owners/admins manage invitations and installation credentials; viewers read reports. Owners manage billing and licences. Upload credentials cannot read reports or manage accounts. Removing a membership denies subsequent access, including from existing sessions; it cannot recall material already downloaded. Disabling publishing stops future uploads, while an owner/admin can separately delete hosted data.

Owners/admins may create named, expiring, read-only agent API keys for their organisation’s hosted findings. We store a hash of each secret, its name, creator, organisation, creation and expiry dates, last-use time and revocation state. The secret is shown once at creation. Keys do not permit uploads, account administration or billing. They stop working after expiry or revocation, the creator’s removal or demotion from admin access, relevant account credential/policy changes, or restored-account quarantine. Agent access can include the same evidence and source excerpts as a hosted report; the customer controls which agents receive keys and how those agents process the data. Key metadata and creation/revocation audit records follow the account/security retention described below.

We process hosted customer report data on the customer’s instructions under the hosted data-processing terms, rather than for marketing or public disclosure. We process account, invitation, licensing, billing and service-security information for the administrative purposes below. Accounts use case-insensitive usernames and salted Argon2id password hashes, plus optional passkeys (public keys and credential metadata, never biometric data). We retain stable internal account IDs, contact email, explicit email proofs, organisation memberships and roles; existing identities are never merged by matching email. To reject compromised passwords, we send only a five-character SHA-1 hash prefix to Have I Been Pwned’s Pwned Passwords range API, with response padding; we never send the password or its complete hash. Password checks fail closed when that service is unavailable. Secure first-party cookies maintain sign-in for 24 hours from sign-in or the most recent authenticated activity (the most recent 20 sessions per account), record authentication method/freshness, and bind five-minute passkey ceremonies. These are essential access cookies, not analytics cookies. Users may require passkeys: password sign-in and email recovery then stop, existing sessions are revoked, and support cannot override the requirement. Another authenticated organisation owner may invite a replacement account; that does not recover the inaccessible account.

Free access for small teams and open-source projects does not require uploading findings or publishing reports. The former automatic open-source disclosure programme has ended. Any retained encrypted submissions are restricted historical evidence and will not be published under that programme. Local reports remain private. Optional private team publishing follows the same user-confirmed data flow described above for free and paid licences; it does not make reports public.

A customer may configure a third-party or local AI runner. The runner is selected and controlled by the customer and may receive source code, prompts and related review context directly from the customer’s environment. Northloom does not receive that material through the runner. The runner provider’s terms and privacy practices apply independently, and customers must assess them before use.

Customers should remove unnecessary personal data and secrets before voluntarily sending support material to Northloom. Any future change to these data flows will be described here before it is introduced.

5. Why we use information and our lawful bases

We use personal data to:

  • assess and respond to enquiries and trial requests (our legitimate interests in developing customer relationships, and contractual necessity where the individual is personally a party);
  • create and administer subscriptions, licences, product access, support and renewals (our legitimate interests in performing and administering business contracts, and contractual necessity where the individual is personally a party);
  • invoice, account for payments and meet tax, corporate and record-keeping duties (contractual necessity and legal obligation);
  • conduct customer, sanctions, export, destination and end-use checks and prevent fraud or misuse (legal obligation where applicable and our legitimate interests in lawful, responsible trade);
  • secure our websites, systems, licences and product, investigate abuse and handle vulnerability reports (legitimate interests and, where applicable, legal obligation);
  • establish, exercise or defend legal claims (legitimate interests and legal obligation); and
  • send relevant business-to-business product communications where permitted (legitimate interests or consent, as required).

Where we rely on legitimate interests, we consider whether those interests are proportionate and whether your rights override them. You may object to processing based on legitimate interests.

Standard Free, paid and trial applications may be checked and approved automatically from the information and confirmations submitted, including registered country, confirmation of use only in supported countries, declared contributor count, prior trial use and any existing block or suspension. Applications that do not pass those checks are held for human review rather than automatically declined. Free open-source applications always receive human review. These decisions concern supply to an organisation; you may contact support@dokima.net to request review or correct an application.

6. Who receives information

We disclose personal data only where reasonably necessary to:

  • service providers supporting website hosting and security, business email and productivity, billing and payments, customer support, software delivery and business administration;
  • professional advisers, auditors, insurers and prospective investors or purchasers under appropriate confidentiality duties;
  • government, law-enforcement, tax, sanctions or regulatory authorities where required or reasonably necessary; and
  • another entity in connection with a reorganisation, financing, sale or transfer of all or part of our business, subject to appropriate safeguards.

Current core providers include Vultr for website and account infrastructure in Manchester, United Kingdom, Backblaze B2 for encrypted recovery storage in Amsterdam, Netherlands, Google Workspace for business email and documents, Stripe for billing and payments, and Resend for transactional verification, recovery, invitation, licence and download email, and Have I Been Pwned for privacy-preserving compromised-password checks. We do not sell personal data.

7. International transfers

Some providers may process personal data outside the United Kingdom. Where UK data-protection law requires a transfer safeguard, we use an applicable UK adequacy regulation, the UK International Data Transfer Agreement, the UK Addendum to approved EU standard contractual clauses, or another lawful mechanism. Contact us for further information about safeguards relevant to your data.

8. Retention

We keep personal data only for as long as reasonably needed for the relevant purpose:

  • unsuccessful sales and trial enquiries: normally 24 months after the last substantive interaction;
  • contracts, customer administration, invoices, payment and screening records: normally for the customer relationship and 6 years afterward;
  • ordinary support records: normally 3 years after closure;
  • reports submitted to us about the security of Dokima or our services: normally 6 years after closure, where a longer record is useful to track product risk and disclosure history; private team reports uploaded at a customer’s request have the separate hosted-data retention described below;
  • website and security logs: normally 30 days, unless an event requires longer investigation or preservation;
  • direct-marketing records: until opt-out or normally 24 months after the last meaningful engagement; and
  • legal disputes or regulatory matters: for as long as required to resolve the matter and meet related legal duties.

Encrypted recovery backups can retain deleted data for years after deletion from the active service. The managed schedule retains snapshots from the most recent 48 hours, plus 30 daily, 12 weekly, 24 monthly and five yearly snapshots. These recovery copies are stored separately in Backblaze B2, are restricted to authorised recovery operations and are not available for ordinary account or report viewing. Deleting active hosted data does not immediately erase encrypted recovery copies. Before restored data becomes accessible, deletion, revocation and current access-policy records must be reconciled and reapplied; otherwise the restored account service stays closed. We may retain anonymised information that no longer identifies an individual.

Private publishing stores the latest complete report per installation and associated discussion history. Questions and answers have no time-based expiry. Inactive installations retain their report and discussion until explicit deletion or withdrawal, subject to documented storage limits, and display their last-upload age. Owners and administrators can delete hosted installations and their active payloads.

Disabling uploads in updated clients requires explicit confirmation that the hosted vulnerability report, Keyholder Report, questions and answers will be deleted immediately. Cancelling leaves publishing enabled. The installation attempts to preserve unsynchronised team contributions locally first. If the service is unreachable, uploads stop immediately, a persisted deletion request is retried, and deletion is shown as pending until confirmed. Enabling uploads again creates a new publication; previous team discussion stays local. Resetting scanning preserves discussion history and does not itself withdraw publishing. Active deletion revokes the publication credential; restricted encrypted recovery copies expire under the managed backup schedule. Invitations and mailbox proofs expire after 24 hours; pairing transactions after ten minutes, passkey ceremonies after five minutes and password reset proofs after 15 minutes; browser sessions after 24 hours without authenticated activity. Expired transaction records are cleaned hourly. Account, membership and audit records are retained while needed to provide and secure the organisation account; contact us to request account/organisation deletion, subject to legal and billing retention duties.

9. Security

We use proportionate technical and organisational measures intended to protect personal data against unauthorised access, loss, alteration and disclosure. No system can be guaranteed completely secure. Please send suspected product or service vulnerabilities to security@northloom.ai and do not include unnecessary personal data.

10. Your rights

Subject to applicable law and exceptions, you may ask us to:

  • provide access to your personal data;
  • correct inaccurate or incomplete data;
  • erase data;
  • restrict processing;
  • provide certain data in a portable format;
  • stop processing based on legitimate interests; or
  • withdraw consent where consent is the lawful basis.

You may opt out of marketing at any time. To exercise a right, contact contact@northloom.ai. We may need to verify your identity and authority. You may complain to the UK Information Commissioner’s Office at ico.org.uk, although we invite you to contact us first so we can try to resolve the issue.

11. Territorial availability

Northloom presently offers Dokima only to approved business customers established in the United States, Canada, the United Kingdom, Australia, New Zealand, Jersey, Guernsey or the Isle of Man. Website controls may restrict access from other locations. Those controls are not a substitute for customer due diligence and may not always identify a visitor’s location accurately.

12. Changes

We may update this notice when our product, providers or legal obligations change. We will post the current version with a revised date and provide additional notice where a change materially affects individuals.