Privacy configuration and legal operations
Prepared 7 November 2026. Applies to the Studio web/API/native clients and each independently operated server. This implementation is not a legal certification. No production settings, notices, mailboxes, retention jobs or firewall rules are changed by merging the source. Missing operator facts remain required work.
1. Publish the selected server's controller information
- Copy
deployment/privacy/privacy-notice.example.jsonoutside the checkout, for example/etc/omi/privacy-notice.json. Keep it operator-owned and readable by the API service. All fields are public; never put credentials, private evidence, unpublished manuscripts or internal connection strings in this file. - Complete the controller's legal name, postal address and monitored privacy
email. Add
dpoEmailonly if a DPO is actually appointed. Identify the competent supervisory authority and its public HTTPS URL. Do not infer the controller from the domain name, project maintainer or hosting provider. - Set
noticeUrlto the installation's complete, accessible HTTPS notice. Add oneprocessingentry per actual purpose: account/authentication, collaborative/review/publishing work, security logs, optional integrations, and communications where applicable. Each needs data categories, a specific legal basis, retention criteria, recipients, and applicable international transfer mechanisms or an explicit statement that there is no such transfer. For legitimate interests identify the actual interest and balancing assessment; for consent explain withdrawal. Do not use consent as a blanket account basis. - Complete rights and complaint instructions, direct/indirect data sources, mandatory/optional fields and consequences of not supplying them, any automated decisions/profiling, cookies/device storage, backups and retained shared data. Mention access, rectification, erasure, restriction, objection, applicable portability and withdrawal where relevant. Cover Art.14 indirect collection separately where invitations/co-authors or third-party metadata make it relevant.
- Publish understandable notice text in the languages of the actual users. UI labels support English, Hungarian and German; configured legal text retains its declared language. The UI does not silently machine-translate legal facts.
- Only after factual review set
operatorReviewed: true, and configureOMI_PRIVACY_CONFIG_PATH=/etc/omi/privacy-notice.jsonin the API environment. Fromserver, runnpm run privacy:checkbefore restarting through the normal deployment procedure. The checker validates shape, not legal compliance. - Verify
/api/v1/privacy,/?privacy=1, registration, sign-in and Account on web and the selected native server. Invalid configured files fail startup with a generic error. An absent file explicitly displays missing information; it does not manufacture a controller or claim compliance. Existing login is preserved. Do not open public registration until the actual notice is complete.
The example deliberately fails validation. No person, address, legal basis, external recipient, DPO, authority, retention duration or operator approval has been invented. Text is rendered as text, not authored HTML; public URLs must be HTTPS without embedded credentials. Notices are not fetched by the server.
2. Receive and fulfil data-subject requests
The Account privacy section and the public notice provide a mailto: link to the
configured controller. It opens the user's mail application; it does not create
a tracked case or send a message by itself. The operator must monitor that
mailbox and offer an accessible alternative for people unable to sign in.
Keep a restricted case register outside public Git/CI with: case ID, received at, request type/scope, verified identity, original one-calendar-month deadline, assigned owner, system inventory, third-party-rights review, decision, response at and final deletion date. Do not ask for an identity document by default. Use proportionate identity verification and avoid collecting unnecessary proof.
Respond without undue delay, normally within one calendar month. Calculate the calendar deadline (not a fixed 30-day timer). Where legally justified by complexity or number of requests, inform the person within the original month of an extension of up to two additional months and its reasons. Refusals need reasons and complaint and judicial-remedy information. Withdrawal of consent must be as easy as giving it.
GET /api/v1/account/data returns only the authenticated user's explicitly selected
profile, additional email addresses, linked sign-in metadata, memberships, session
metadata and passkey labels. It returns no password hash, token, key or identity
provider raw profile. The download is private/no-store and rate-limited. Sections
over 1,000 rows are explicitly marked incomplete. This account overview is not
a complete Art.15 response, complete Art.20 package or manuscript backup.
For full access search both Identity and Studio stores, collaboration/review history, submitted versions, service logs, support cases and relevant processors. Review shared content to protect other people's rights and reviewer confidentiality; provide a reasoned response rather than automatically exporting every shared record. For portability check the consent/contract and automated-processing conditions, then supply the applicable user-provided/observed data in a usable machine-readable format. Use the normal manuscript exports where appropriate. Restriction and objection need an actual processing decision; merely disabling login is insufficient.
3. Erasure, retained shared history and restoration
The existing self-service deletion removes live Identity data and credentials, clears account integration records and pseudonymizes the Studio participant. A stable participant ID or identifiable shared content can remain personal data. Pseudonymization is not anonymization, and scholarly provenance alone is not blanket permission for indefinite retention. Review the full record, lawful basis, necessary scope and retention; assess research/publication exceptions individually.
The last-owner safety block must not indefinitely defeat an erasure request. Offer a verified operator-assisted handover/alternative within the same request process. Record and explain any lawful partial refusal. Review publication venues, shared annotations, signatures, audit JSON, profile snapshots, object storage, queues/caches and previous exports as well as the primary user table.
Keep a minimal, restricted erasure/restriction register for the approved backup window. A restored backup must remain isolated until deletions/restrictions are reapplied and obsolete sessions/credentials invalidated, before normal traffic or external synchronization. Test this with synthetic accounts. The register itself contains personal data and needs a purpose, access control and expiry.
The software cannot enforce a VPS backup rotation. Inventory all replicas, snapshots and offsite/provider copies, set a justified maximum age, restrict access, and test expiry plus recovery. Honour any previously published 30-day commitment; do not silently extend it through this generic notice update. External OJS/OMP or user-controlled cloud copies need the responsible destination's own procedure.
4. Retention and expiry cleanup
Maintain a restricted schedule for each store: purpose, legal basis, data owner, retention trigger, maximum age/criteria, deletion/anonymization method, legal hold, processor propagation and evidence of an executed job. Include access/auth logs, review history, invitations, OAuth states, exports, support cases, audit events and backups. No universal statutory log-retention period is assumed.
The Studio implements technical audit retention defaults, independent of
systemd/Plesk log rotation: 90 days for integration operation digests
(integration_audit_events) and module operation logs
(studio_module_audit_events); 180 days for central/institutional administrator
events (admin_audit_events) and collaboration access/role changes
(collaboration_audit_events). These are proposed project defaults, not a
universal statutory deadline. Each controller must record its rationale,
lawful basis, access controls and any justified preservation of incident or
legal evidence. The generic cleanup excludes manuscripts, review/editorial
evidence and user-authored translation variants; handle these categories
through separate retention decisions.
From server, npm run privacy:cleanup remains dry-run and reports
expired authentication records and technical audit candidates, thresholds
and counts without user identifiers or digests. After review,
npm run privacy:cleanup -- --apply removes expired authentication records
and the specified aged technical audit classes in bounded batches.
Failures may leave partial progress; the task can safely be rerun and
reports any backlog. The code does not install a scheduler: the operator
must run a migration, test the dry-run and restore procedure, review legal
preservation duties, then explicitly enable the provided systemd
service/timer examples. Regular execution must be monitored.
See the Studio operator guide
and the systemd templates in deployment/systemd/.
Do not claim that automated purging runs on a live server until its
operator confirms the enabled timer.
Central OMI Studio: observed operational activation (16 November 2026)
For the official central service only: on 16 November 2026, the
operator provided production terminal evidence that
omi-privacy-cleanup.service completed a manual --apply run
successfully at 12:20:49 UTC. The report had
candidates: 0, removed: 0 and remaining: 0 in all
four technical-audit categories, with backlog: false;
expired authentication credentials were also processed.
The daily omi-privacy-cleanup.timer was verified
enabled and active and showed the next run at
17 November 2026, 03:24:13 UTC (04:24:13 CET).
The screenshot showed LAST=-: the first unattended
timer execution had not yet occurred at the verification
time. Do not describe it as already completed.
The restricted-scope public operational evidence record distinguishes the confirmed manual run and schedule from pending post-activation checks. The source is operator-supplied screenshots, not independent SSH access or a tested DB restore. Backup restore verification, active controller notice wording and subsequent successful unattended runs still require follow-up. Institutional and third-party deployments are independent and do not inherit central server retention settings.
5. Personal-data breach procedure
Assign an incident owner, deputy and monitored security/privacy contacts. Detect, contain, preserve restricted evidence, record when the organization became aware, and assess confidentiality, integrity and availability impacts on people. Register every personal-data breach, including the reasons for not notifying.
The processor informs the controller without undue delay. A controller notifies the competent supervisory authority without undue delay and, where feasible, within 72 hours of awareness unless the breach is unlikely to risk individuals' rights/freedoms. Explain delays and provide information in phases if needed. Inform affected people without undue delay where the breach is likely to create high risk, subject to the legal exceptions. Do not automatically send private logs or manuscripts to GitHub, providers, authorities or users.
Restricted incident record: ID; discovery/awareness UTC; reporter/owner; affected systems and data categories; approximate people/records; likely consequences; containment and recovery; risk and notification decisions/reasons; authority and individual notifications/times; processor notices; corrective actions; review date. Use synthetic tabletop and restore exercises to test responsibilities and timing.
6. Institutions, providers and international transfers
Map controllers/processors/joint controllers per activity. A software maintainer is not automatically controller for an institution's independently hosted data. Where hosting/processing is on an institution's behalf, conclude the Art.28 agreement before processing: scope, duration, data/people, instructions, confidentiality, security, subprocessor authorization/change notification, rights/incident assistance, deletion/return, audits and international transfers.
Inventory actual hosting, mail, OJS/OMP, Office, identity, storage, translation, reference/research and AI providers, including locations, recipients, subprocessors and enabled data flows. Record applicable adequacy decisions or safeguards and required transfer assessments; user initiation of a connection does not waive these obligations. Disable unused integrations and minimize transmitted content. Perform DPIA screening before high-risk deployments, especially sensitive or large-scale research data. Assess Art.9 conditions where special-category data is involved, Art.30 records and Art.37 DPO criteria; do not declare them universal or automatically exempt a small provider from recurring processing records.
7. Cookies and device storage
Inventory cookies, local/session storage and SDK/network behavior on web/native and the documentation site. The current change adds no tracking, analytics or consent checkbox. Classify by actual purpose; not every local-storage use is exempt. If all storage is strictly necessary for a requested service, explain it without an unnecessary consent banner. Before adding optional tracking, implement prior purpose-specific opt-in, equally accessible refusal, withdrawal and a minimal consent record. Prevent the optional technology from running before choice.
8. Conditional regulatory register
Keep each decision with legal entity, deployment/service scope, evidence, responsible owner and next review date. No eligibility decision is made by code.
| Regime | Required determination and resulting work |
|---|---|
| Hungarian cybersecurity law / NIS2 | Assess the actual operator's sector, size and designation under Act LXIX of 2024; covered operators need the applicable registration, risk/security, incident and audit duties. Software use alone does not determine scope. |
| Cyber Resilience Act | Assess manufacturer/commercial distribution versus noncommercial FOSS or qualifying steward roles and remote-processing scope. In-scope manufacturer Art.14 reporting applies from 11 September 2026 (24h early warning / 72h notification); the main obligations apply from 11 December 2027. Steward Art.24(3) reporting starts 11 December 2027. Prepare vulnerability handling, supported-update policy, technical documentation/SBOM, required conformity evidence and reporting only for the applicable role/scope. No CE/compliance claim is created here. |
| Accessibility | Assess public-sector operator obligations under Act LXXV of 2018, and separately EAA-covered services/products and exemptions. Where applicable provide tested keyboard/screen-reader access, an accurate accessibility statement, feedback and enforcement contacts. Automated UI tests alone are insufficient. |
Before declaring compliance, fill the controller notice, assign real staff, activate and verify actual retention/backup processes, sign necessary agreements and obtain the relevant role/scope assessments. A source merge is not that evidence.
Official references
- GDPR: Arts.5–6, 9, 12–22, 25, 28, 30, 32–37, 44–49 and 89.
- EDPB: individual rights.
- EDPB: cookie consent.
- Commission: breaches.
- CRA open source, reporting, implementation.
- Hungarian cybersecurity act.
- Hungarian public-sector accessibility act.
- EAA scope.
Privacy API compatibility checkpoint — 2026-11-08
New clients use the v1 URLs above. /api/privacy and /api/account/data remain
shared-handler legacy facades with deprecation signals and a common quota. Deploy
the repaired API before clients. The scoped omi-account-overview/1 body is not
a complete access/portability response, and no controller facts are invented.
See the architecture reconciliation
and Studio ADR-034 for schemas, errors, repository boundaries and release evidence.
Status is proposed until companion source merge/CI; operator acceptance is separate.