Privacy Notice

Authenticated public knowledge and host-context privacy notice

What Legal-MT Is

Legal-MT is a free public knowledge and tool plugin/MCP for Malta and applicable EU legal sources, procedures, editable templates and workflow assistance in Codex and ChatGPT. All publisher legal/source/template resources are public; a historical operational label does not make them client-confidential documents. There is no client-facing web workspace.

Public Queries And Host Context

Public tools process submitted search terms, source IDs, page/article selectors, minimal issue summaries, direct structured urgency answers and workflow/resource filters to return knowledge or compute a response. The read-only knowledge tools do not create matter records, uploaded documents, drafts or case histories. Sign-in, account status and OAuth/security records are separate operational data described below. Do not submit attachments or identifying case narratives to these tools. User context, selected files, reasoning, generated artifacts and continuity are handled by the user's Codex or ChatGPT host and its available tools.

Purpose And Data Minimization

Tool arguments are used to answer the requested public knowledge or routing query. Submit the minimum necessary public query, not personal case files, identity or medical records, passwords, API keys or credentials. Public routing is not verified legal classification, a medical classifier, conflict clearance or a grant to access another service. Host-specific processing and disclosure remain subject to the host's policies and the user's authority for the selected context.

Public Knowledge Storage

The operator hosts public source/catalog/provenance metadata, editable templates and available public corpus objects. Source hashes, versions and URLs support integrity and traceability; they do not certify current law or legal merits. The public plugin does not maintain a separate client-case database or save user artifacts in an operator work-product bucket. Host-created files are not remote Legal-MT saves.

Authentication And Operational Security

Access to all eight public knowledge tools requires OAuth through Google sign-in or a verified email account and administrator activation. Account identity, access status, onboarding, registered OAuth clients, authorization and token records, security events and support records are operational data, separate from public legal resources and host-owned user files. Authentication does not require a conflict-cleared matter, file upload, server artifact save or reviewer appointment. Credentials, tokens, keys, backups and administrative access controls must not be published with legal resources.

Hosting And External Services

Cloudflare hosts the public Worker/MCP and its public source storage and may process request, network and security metadata under its applicable settings and policies. OpenAI provides the Codex or ChatGPT host; host attachments, conversations and generated files follow its applicable policies and tool behavior. Google sign-in, verified-email authentication and operational contact services process necessary identity, authentication or contact metadata; they are not client-file upload or case-storage services of the public plugin. Legal-MT does not require a second plugin, external model API or user-supplied model API key.

Public Query Retention

Public knowledge and routing queries are separate from sign-in, account-status and OAuth/security records. The eight public knowledge tools do not upload or save user files, create client-case records or call a separate model API. Operational records and infrastructure request logs may be retained separately, as described below. Codex, ChatGPT and connected host tools have their own retention policies.

Automatic Security-Record Cleanup

Authentication and operational functions retain security records. A daily cleanup is scheduled for 02:17 UTC. It selects OAuth state, authorization-code and one-time-code challenge rows once their expiry or recorded consumption/use is more than one day old; rate-limit buckets once their start is more than two days old; access-token records once expiry or revocation is more than 30 days old; and refresh-token records once expiry, revocation or rotation is more than 30 days old. Deletion takes place on a successful cleanup run after the threshold, not at credential expiry. Failed or delayed runs can retain rows longer. These rules do not delete the separate authorization-code linkage records.

Operational Records Retained Until Deletion

Account and identity metadata, access status and onboarding records, registered OAuth clients, authorization-code linkage records, administrator and historical invocation audit records, contact/support messages and data-rights requests have no automatic age-based deletion in the current application. They remain stored indefinitely until separately deleted by the operator. Public source and provenance records are also retained without a fixed application expiry. Disabling an account, revoking a token or acknowledging a deletion request does not erase these records. The account deletion endpoint records a request for review; it does not perform automatic erasure.

Cloudflare Logs And Recovery Copies

As checked on 9 October 2026, persisted Cloudflare Worker invocation logs are enabled at 10% sampling. Request metadata is logged; URL query-string redaction was enabled and verified on that date. This change does not establish deletion of older log entries, which remain subject to their retention window. Worker Logpush and tail consumers are not configured, and tracing is disabled. Cloudflare documents a seven-day Workers Logs retention window for the paid Standard usage model shown in this deployment, and up to 30 days of automatic D1 database recovery history. Deleting a live database row therefore does not immediately remove recoverable earlier copies. Separately created operator backup copies do not inherit that 30-day expiry: the application does not automatically purge them, and they remain until separately deleted. These windows do not cover every Cloudflare security or account record.

Separate Administrator-Access Logs

Cloudflare Access protects administration and has separate retention rules. The account dashboard checked on 10 October 2026 (Africa/Kigali) showed the Zero Trust Free plan. Cloudflare documents 24-hour retention for Access authentication logs on that plan and 18 months for Zero Trust administrator logs. The account-scoped Logpush listing showed no configured jobs or export destinations at that check. This does not establish that historical or manually exported copies are absent. Log Explorer datasets and their retention remain unverified, so no single deletion date covering all copies is promised. These logs are separate from the application's administrator audit records, which remain until deleted.

Codex And ChatGPT Retention And Rights

OpenAI controls host conversations and files under the applicable product and workspace settings. Its published ChatGPT policy keeps regular and archived chats until deletion or an applicable workspace policy removes them, then schedules deletion within 30 days, subject to de-identification and legal/security exceptions. Saved Library files require separate deletion. OpenAI also documents a 30-day deletion process with stated exceptions for Codex chats in the ChatGPT app. This does not establish deletion of files saved on your device or in connected services. Use the relevant host and file-store controls. For operator-held records, contact info@ikanisa.com; Legal-MT cannot delete host-held data and does not treat a received request as completed erasure.

Contact And Operational Records

Privacy, correction, access or deletion questions about operator-held support or authentication/security metadata can be raised through the published support channel. Where applicable account metadata endpoints remain supported, /account/export and /account/deletion-request concern those operational records, not a Legal-MT case-file store or host files. Identity and authority checks may be required before disclosing or deleting operator-held records. Public research requires neither endpoint.

Transfers And Security

Cloudflare hosting, OpenAI host processing and separately used operational services may involve international processing. Applicable provider terms, settings and transfer arrangements govern their own processing; no universal location, zero-retention or absolute-security assurance is made here. Restricted operator administration remains separate from public knowledge access.

Who Can Use Legal-MT

The public library supports individuals, professionals, businesses and institutions researching Malta and applicable EU issues through authenticated access and an administrator-controlled active account. Use remains subject to the host's applicable age and account terms; Legal-MT does not introduce professional appointment or professional approval prerequisites for ordinary research and preparation.

Professional boundary

Public Malta and EU knowledge, sources, templates and workflow assistance through Codex and ChatGPT. The host handles user context, reasoning and generated files; Legal-MT does not maintain a client case-file store or receive user attachments. No client-facing web portal, certified legal opinion, filing, authority contact, STR/SAR, payment processing or external delivery.

Policy pack

Delivery surface: Public Codex and ChatGPT plugin/MCP; no public web user portal. Public web user portal: none.