Donghyeok Kim (김동혁, the "Operator") establishes and discloses this privacy policy under Article 30 of the Personal Information Protection Act of the Republic of Korea (the "Act"), to protect the personal information of users of WikiDrive (the "Service") and to handle related complaints quickly and smoothly. The Service includes the website (https://wikidrive.app), the AI connection address (https://wikidrive.app/mcp), and the plugins for Claude Code and ChatGPT that the Operator distributes. Matters about the former SEED service are governed by Article 16.
Note: the draft of this privacy policy was written with AI (Claude). This English version is a reference translation; if it differs from the Korean text, the Korean text prevails.
(1) The Operator processes personal information for the following purposes. It does not use it for any other purpose; if a purpose changes, it takes the required measures, such as obtaining separate consent under Article 18 of the Act.
(2) The Operator processes the personal information in paragraph (1) under Article 15(1)4 of the Act to perform the service agreement made with the member, and does not ask for separate consent for this. Where the table in Article 2(1) states a different basis for an item, that basis applies, and information sent to a connected AI service is governed by Article 6(2).
(3) The Operator does not send advertising messages, and does not sell members' personal information or content or use it for advertising or for training AI models. Whether an AI service the member connected uses the information it receives for training is, however, governed by that service's policies and the member's settings there.
(4) The Operator does not process pseudonymised information under Article 28-2 of the Act and does not make additional use or provision under Articles 15(3) and 17(4) of the Act.
(1) The Operator processes the following items. In the table, (Required) marks items needed to sign up or to use that feature, (Generated) marks items created when the Service is used, and (If applicable) marks items processed only when that feature is used.
| Category | Items | How collected | Basis |
|---|---|---|---|
| Account | (Required) e-mail address, time the e-mail was confirmed, last sign-in time, account ID, interface language, how the account was created (web, AI connection, or carried over from the former service) (If applicable) Signing in with Google: the account identifier, e-mail address and whether it is verified, name and profile picture address that Google provides (the Service uses only the name, to suggest a display name) (If applicable) Accounts that signed in to the former SEED service with GitHub: that link (If applicable) Accounts that have a password: the password hash (the Service does not take a password from new members; only accounts that already have a password can sign in with it) | Member, Google | Contract |
| Sign-up consent record | (Generated) Time and screen (sign-up, sign-in, AI connection approval, etc.) at which the member agreed to the Terms of Service and confirmed being 14 or over, and the editions of the Terms of Service and privacy policy at that time | Generated automatically | Contract |
| Sign-in session | (Generated) IP address, browser information, times the session was created and refreshed | Generated automatically | Contract |
| Profile | (Required when first using the website) name (1 to 40 characters), status (one of: undergraduate, graduate student, researcher, faculty, employee, founder, other), field (one of 9) | Member | Contract |
| Settings | (Generated) Whether saving is paused for all AIs, the one-line AI notice, whether a repeated correction is marked as confirmed, whether automatic documents are on or off and when that was chosen, when the automatic-documents notice was shown for each kind of AI, when notifications were last seen | Member's choice, generated automatically | Contract |
| Content | (If applicable) Titles and bodies of documents and prompts, topics, folder names, rule texts and their states, workflow names and steps, attachments (file name, type, size and contents), the title and body at every change (change history), whether the author was the member or an AI and which kind of AI, whether a document was made by an AI and whether on request or automatically | Member, sent by the connected AI | Contract |
| Sharing and groups | (If applicable) The member who shared, invitees' e-mail addresses (including people with no account), role (viewer, editor), status (pending, accepted, declined, revoked), whether AI use is on (it is turned on when a share is accepted), the link between a shared copy and its original and how changes are carried over, group names and members | Entered by the sharing member, generated automatically | Members: contract Invitees who are not members: legitimate interest of the controller (Article 15(1)6 of the Act) |
| Web publishing | (If applicable) A copy of the title, body, rules and topic name at the time of publishing, publishing time, the publishing member's ID (never shown) | Member's choice | Contract |
| AI connections | (If applicable) The name and return address the requesting AI service registered, the kind of AI derived from that address (ChatGPT, Claude, a program on the member's computer, other), member and connection IDs, permission (read, or read and write), times connected, expiring, last used and revoked, whether saving is paused for the connection, authorization request data, hashes of authorization codes and access and refresh tokens, a first-use mark per kind of AI | Generated automatically | Contract |
| Requests the AI sends | (If applicable) Topic (up to 40 characters) and topic name, recall keywords and search queries (not stored), rule texts, document titles, bodies and headings, reason for saving (asked or automatic), request key | Sent by the connected AI | Contract |
| Usage counts | (Generated) Counts per member, kind of AI and day (UTC), per document and day, and per connection and day (loads, opens, searches, rules saved, repeated, confirmed and changed, prompts saved, documents created, edited, undone and refused, etc.); per document and kind of AI, the running total of loads and the last time; activity records (who did what through which connection, and how many) | Generated automatically | Contract |
| Stability records | (Generated) Duplicate-processing records (hash of the request key, hash computed over the whole request, IDs of the result) Web request records (the request ID and the ID of the document created) Web write request records (the request ID and the content of that request: the topic, title and body of a document as first typed, invited e-mail addresses, folder and group names, profile, attachment name, type and size, etc.) Times of web writes (for rate limits) The network address that sent an AI connection registration (the IPv4 address, or the first 64 bits of the IPv6 address) converted with a key held only on the server (the original address is not stored), and counts per hour Storage paths of attachments waiting to be deleted and the ID of the member who asked for the deletion | Generated automatically | Contract |
| Inquiries | (If applicable) E-mail address, name (if given), content of the inquiry or report and attachments | Sent by the person who asks | Members: contract Non-members: legitimate interest of the controller (Article 15(1)6 of the Act) |
| People introduced on the service introduction page | (If applicable) Name, affiliation or status, mark (Prompt Engineer, etc.) | From the person directly | Consent (Article 15(1)1 of the Act) |
| Deletion records | (Generated) The deleted member's ID, time and method of deletion | Generated automatically | Legitimate interest of the controller (Article 15(1)6 of the Act; confirming the deletion) |
| Access logs | (Generated) IP address, time and address of requests, browser information (logs that processors keep while delivering the Service) | Generated automatically | Members: contract Visitors not signed in: legitimate interest of the controller (Article 15(1)6 of the Act) |
| Former SEED service | (Generated) The basic profile of the former SEED service created together with each account (display name "SEED 이용자") and the information in Article 16 | Generated automatically | Contract (Article 16) |
(2) If a document, prompt or rule sent by a connected AI, or a rule a member types on the website, contains an e-mail address, a phone number, a number in the format of a Korean resident registration number, a long run of digits, a passport number, an IBAN, a value that looks like a password, key or token, or a webhook address, the Service does not store it (for documents sent by an AI, it counts the refusal). Rules and prompts sent by an AI are also not stored if they contain a web address. This check only looks for set patterns, so it does not catch all personal information, and it does not apply to the bodies of documents and prompts a member writes on the website, or to attachments.
(3) Members must take care not to put sensitive information (health, beliefs, etc.), unique identifiers such as resident registration numbers, passwords or keys, or card or account numbers into their content, and may put other people's personal information into it only as far as they are entitled to. The Operator does not collect such information on purpose, but if it is in content a member enters or an AI sends, it is stored as part of that content.
(4) Attachments go straight from the member's browser to the processor's storage, and the Operator does not inspect their contents. Attachments are not passed to AIs through AI connections.
(5) The Service uses recall keywords and search queries only to answer the request and does not store them; for searches it only counts. It also does not store the program name and version an AI service sends each time it starts a connection. The name and return address sent when the connection is first registered are stored as the "AI connections" item in paragraph (1).
(6) Usage counts contain no conversation content, no document content and no keywords or queries. They are, however, stored with the ID of a member account, document or connection, so they are not anonymous.
(7) The Service's servers do not receive the conversation history of connected AI services. They only receive the values a connected AI sends when it calls a function (tool) of the Service. The Claude Code plugin looks at the current message on the member's computer as described in Article 11(6). However, while automatic documents are on (Article 12(3)), the AI may send content summarised from the conversation as a document without the member asking each time.
(8) Invitees' e-mail addresses are entered by the sharing member. If an invitee asks, the Operator tells them, under Article 20 of the Act, the source of the information, the purpose of processing and that they have the right to demand that processing stop. An invitee who is not a member may demand by e-mail (support@wikidrive.app) that processing of their e-mail address stop and that it be deleted.
(1) The Operator does not accept sign-ups from children under 14.
(2) Before an account is first used, the Operator asks the person to confirm on an agreement screen (the sign-up screen, the AI connection approval screen, or the terms and privacy screen shown after signing in) that they are 14 or over, and if it learns that personal information of a child under 14 has been collected, it destroys that account and information without delay.
(1) The Operator processes and keeps personal information for the periods in paragraph (2). The Operator currently keeps no personal information separately because another law requires it; if it comes to do so, it will state the basis and items in this policy.
(2) Retention periods are as follows.
| Category | Retention |
|---|---|
| Account, profile, settings, sign-up consent record | Until the member deletes the account |
| E-mail address of an account that requested a sign-in code but did not finish signing up | Becomes eligible for deletion if sign-up is not finished within 7 days (only accounts that have not yet signed in with the code; an account that has signed in follows the next row) |
| Account that has signed in but has not agreed to the terms and confirmed being 14 or over | Until the person deletes the account or asks for its deletion by e-mail (it is not deleted automatically) |
| Sign-in session | Until account deletion (signing out ends that session; how long the processor keeps records of ended sessions follows the processor's policy) |
| Documents, prompts, rules, topics, folders, workflows, attachments | Until the member deletes them or the account |
| Shared copies | Until 30 days after the member moves that copy to the trash, or until the member deletes the account (they stay even if the original is moved to the trash or deleted) |
| Documents in the trash | Become eligible for deletion, with their change history, rules and attachment records, 30 days after being moved to the trash |
| Folders in the trash and the documents in them | Become eligible for deletion 30 days after being moved to the trash. This deletion starts on 24 November 2026 (folders moved to the trash before then become eligible from that day), and the cleanup removes a large folder a set number of rows at a time over several runs (the member can delete them forever from the trash before then) |
| Change history | Until the document is deleted (earlier text stays even after edits and undo) |
| Texts of rules taken out of the current document | Until the member erases that rule or the document is deleted |
| A new rule an AI saved and then undid within 24 hours | Deleted at the undo, with no history kept |
| Sharing records (including invited e-mail addresses) | Until the shared item or space is deleted or the sharing member deletes the account. The e-mail address in a record of a revoked share or a declined invitation, however, becomes eligible for deletion 30 days after that day. When a person who accepted deletes their account, the records of the shares they accepted are deleted, but invitations not accepted stay even if the invitee deletes their account |
| Groups | Until the member who created the group deletes the account (there is no way to delete a group itself; documents in it can be moved to the trash and deleted) |
| Published web copies | Deleted at once when publishing stops. Moving the document to the trash or to another space ends publishing, and restoring it does not publish it again |
| Web write request records (including the request content) | Become eligible for deletion 30 days after creation (deleted earlier if the member deletes the account or that space is deleted) |
| Web request records | Become eligible for deletion 30 days after creation |
| Duplicate-processing records | Become eligible for deletion 30 days after creation |
| Times of web writes | Deleted at the member's next write after 1 hour, or by cleanup after 1 day |
| Usage counts per member and per document | Days before the latest 365 days (including today, UTC) become eligible for deletion. Member records are deleted with the account, document records with the document |
| Usage counts per connection | Become eligible for deletion 365 days after that day, or when the connection is deleted |
| Running load totals per document | Until the document is deleted |
| First-use marks per kind of AI | Until the member deletes the account |
| Activity records | Become eligible for deletion 365 days after creation |
| AI connections | Expire 90 days after last use. A revoked or expired connection becomes eligible for deletion, with its token hashes, 1 day later |
| Authorization requests | Valid for 10 minutes (the code for 5 minutes); eligible for deletion 1 day after expiry |
| AI connection registrations (no member ID is stored in them) | A registration never approved is deleted at the next registration after 7 days. An approved registration is kept with no time limit, so the same AI service can connect again |
| Registration rate-limit records | Records of past hours are deleted in turn at later registrations |
| Deletion records | Become eligible for deletion 30 days after the account was deleted |
| Inquiry records | 1 year after handling is finished |
| People introduced on the service introduction page | Until the person withdraws consent or the introduction ends |
| Access logs (processors) | According to the processor's policy |
| Information of the former SEED service | Article 16 |
(3) Records that "become eligible for deletion" in paragraph (2) are removed by the Service's cleanup. Cleanup runs at most once an hour, only when the Service is used (for example when the website is opened or an AI loads a prompt), and removes a set number of rows per run, so actual deletion may happen later than that point. The cleanup of usage counts runs when an AI loads a prompt or views usage, or when the "Usage" screen in Settings is opened.
(4) When a member deletes their account, the Operator deletes at once: the account; all content in the personal drive (including change history, rules, attachments, published web copies and workflows); shared copies the member made; groups the member owns and their content (also for the other members); shares the member gave and shares the member accepted; AI connections and tokens; the member's web write request records; duplicate-processing records; usage counts per member; the profile and settings. The following remain for the stated time:
(1) When personal information is no longer needed, for example because the retention period has passed or the purpose has been achieved, the Operator destroys it without delay.
(2) Information in electronic form is deleted from the database and storage, and deleted information cannot be brought back in the Service. Attachment contents and copies kept by processors are governed by paragraphs (3) and (4).
(3) When an attachment is deleted, its contents are put on a deletion queue, and from then on no one but the Operator can open them. Queued files are removed from storage by the member's browser when that member opens a Service screen, and the files of a member who deletes their account are removed from storage at the deletion by that member's browser, in the member's own signed-in session. Files that could not be removed then are removed by the Operator after checking. File paths, which consist only of random IDs, are kept so that the same path is never used again.
(4) The Operator does not make or keep backup copies of its own. Any copies a processor keeps against outages are deleted under that processor's policy.
(1) The Operator does not provide personal information to third parties except in cases under Articles 17 and 18 of the Act, such as the data subject's consent or a special provision of law.
(2) When a member starts a WikiDrive connection from an AI service and clicks "Allow" on the approval screen, the Service sends the following to that AI service when it calls the Service's functions. The Operator makes this provision on the basis of the member's consent given on the approval screen (Articles 17(1)1 and 28-8(1)1 of the Act), and the approval screen states the recipient (with its country and contact, where the Operator can know them), the items, the purpose, when and how the information is sent, the retention, and the right to refuse consent and how to refuse. The member may refuse consent; if they do, they cannot use AI connections, but can still use the Service on the website.
| Recipient | Operator and country | Items | Purpose | Retention |
|---|---|---|---|---|
| The AI service the member connected (e.g. Claude, ChatGPT) A program on the member's computer (e.g. Claude Code, which sends what it receives to its own AI provider) | Claude (including Claude Code): Anthropic PBC, United States (privacy@anthropic.com) ChatGPT: OpenAI OpCo, LLC, United States (privacy@openai.com) Members living in Europe, and other AI services and programs: the provider of the AI service the member connects. The Operator cannot identify that provider or its country, so the member checks them in that service | 1. Titles, bodies, rules and rule states, versions and headings of the member's prompts and documents, and the IDs and topic keys of documents and rules 2. The member's topic list, workflow names and steps, usage counts 3. Names and part of the content of shared items and groups the member accepted, except those for which the member turned AI use off (AI use is turned on when a share is accepted; up to 3 at a time, up to 1,500 characters each) 4. Whether saving is paused, whether automatic documents are on, interface language 5. Results of saving documents and rules the AI made or changed | To let the member apply their prompts and rules and manage their documents in that AI | According to that AI service's policies and the member's relationship with it |
(3) The information in paragraph (2) is sent over the network (HTTPS) to that AI service each time the connected AI calls a function of the Service. If the recipient is abroad, this is a transfer abroad (Article 8(3)); the member can refuse it by not clicking "Allow" on the approval screen or by disconnecting under paragraph (5).
(4) Through a connection under paragraph (2), the Service does not attach the member's e-mail address or name, other members' e-mail addresses, or attachments. What is written in the text of documents and rules is, however, passed on as it is. A connection for which the AI service asked only for read permission cannot save rules or documents.
(5) AI connections are optional; the website can be used without one. A member can disconnect, and so withdraw consent, at any time under "Connected AIs" in Settings, after which access is blocked. Information already passed to an AI service is kept, used and deleted under that service's policies, so the member should also ask that service to delete it if needed.
(6) When a member shares content with a person or group:
(7) When a member publishes a document to the web, anyone, including people not signed in, can view the published copy (title, body, rules, topic name) at its address, and the copy also appears in the public list, newest first. Published pages do not show the publishing member's name or e-mail address, but what is written in the document is published as it is. Published pages are marked to keep search engines from collecting them, but copies others have already made cannot be recalled.
(1) The Operator entrusts processing as follows to provide the Service smoothly.
| Processor | Entrusted work |
|---|---|
| Supabase Pte. Ltd. | Running the database (all information in Article 2), member authentication (including handling sign-in code e-mail requests), storing attachments |
| Cloudflare, Inc. | Hosting and delivering the website, processing and relaying AI connection requests, running the domain, forwarding e-mail sent to wikidrive.app addresses to the Operator's mailbox |
| Plus Five Five, Inc. (Resend) | Sending sign-up and sign-in code e-mails |
(2) If the entrusted work or a processor changes, the Operator discloses it without delay through this privacy policy.
(3) Companies to which a processor in turn entrusts work (sub-processors) are those in the following lists each processor publishes.
(4) When a member signs in with a Google account, the sign-in button and window are provided directly by Google LLC, and what Google processes is governed by Google's privacy policy. As a result, the Operator receives the account items in Article 2(1).
(1) Under Article 28-8(1)3(a) of the Act, to perform its contract with members, the Operator entrusts processing and storage of personal information abroad as follows, including where personal information is accessed from abroad.
| Recipient (contact) | Country | When and how | Items | Purpose | Retention |
|---|---|---|---|---|---|
| Supabase Pte. Ltd. (privacy@supabase.com) | Singapore company (database stored in Seoul, Republic of Korea. The region of attachment storage and the places from which Supabase accesses data from abroad follow Supabase's policy) | Over the network, when Supabase accesses it from abroad to run and support the service | Personal information in Article 2 and attachments | Work in Article 7 | The periods in Article 4, or until the processing contract ends |
| Cloudflare, Inc. (privacyquestions@cloudflare.com) | United States and countries where the Cloudflare network is present (requests are handled near the user) | Sent over the network each time the website is visited or an AI calls the Service | IP address, request address, time and browser information, information in requests and responses (including documents and rules an AI sends or receives), forwarded e-mail | Work in Article 7 | According to Cloudflare's policy |
| Plus Five Five, Inc. (Resend) (support@resend.com) | United States | Sent over the network when a sign-in code e-mail is sent | Recipient e-mail address, sign-in code | Sending e-mail | According to Resend's policy |
(2) A data subject can refuse transfer abroad by not visiting the website, by not signing up, or by deleting their account. In that case the Service cannot be used. An invitee who is not a member may demand by e-mail (support@wikidrive.app) that processing of their e-mail address stop and that it be deleted.
(3) Information sent to the AI service a member connected under Article 6(2) may be transferred abroad as follows, on the basis of the member's consent given on the approval screen (Article 28-8(1)1 of the Act).
Under Article 30 of the Enforcement Decree of the Act, the Operator takes the following measures.
(1) Sensitive information a member writes in content can be seen by others if the member shares that content or publishes it to the web.
(2) All content is stored privately; sharing and web publishing happen only when the member chooses them. Before publishing, the publishing screen states what will be published and that anyone can see it. A member can check recipients and roles on the sharing screen and can make content private again at any time by revoking shares or stopping publication. Copies already saved by people who viewed it cannot be recalled.
(3) The Operator shows in advance, on the sharing and publishing screens, a notice that sensitive information may become visible to others.
(1) The Service (WikiDrive) does not use cookies or tracking tools for analytics or advertising, and does not collect behavioural information for targeted advertising. Analytics on the screens of the former SEED service are governed by Article 16.
(2) The Service uses browser storage and cookies as follows.
| Where | Contents | Purpose | Kept |
|---|---|---|---|
| Local storage | Sign-in session (access and refresh tokens, account information) | Staying signed in | Until sign-out |
| Local storage | Interface language, list or grid view | Display settings | Until the user clears it |
| Session storage | Unsaved text being edited, new-document requests (title, body), share attempts (e-mail addresses), retry records for folder, move and bulk actions | Not losing work if the connection drops | Until the browser tab is closed |
| Cookie (wd_consent) | A random value (the server stores only its hash) | Protecting the AI connection approval screen | 10 minutes |
(3) The Google sign-in script is loaded only on sign-in and sign-up screens and in the sign-in box of the AI connection approval screen opened while not signed in, and Google may then use its own cookies, under Google's policies.
(4) Because of settings Cloudflare, a processor, adds to responses, the user's browser may send connection error reports to Cloudflare.
(5) Users can delete stored information or refuse storage in their browser settings. Sign-in and AI connection approval may then not work.
(6) If a member installs the Claude Code plugin, it works on the member's computer as follows.
(1) A data subject may at any time exercise rights against the Operator, such as access, correction or deletion, demanding that processing stop, and withdrawing consent.
(2) Members can do the following themselves on Service screens.
(3) Automatic documents work as follows.
(4) Requests that cannot be done under paragraph (2) (changing the e-mail address, access to information not in the download, deleting a single version, deleting a group, deleting web write request records, confirming deletion of attachment contents, account deletions that need the Operator, requests from invitees, etc.) can be made by e-mail (support@wikidrive.app). The Operator acts within 10 days of receiving a request, reports the result, and charges no fee.
(5) The rights in paragraph (1) may be exercised through a representative such as the data subject's legal representative or a person given a power of attorney, who must submit the power of attorney.
(6) If the Operator refuses a demand or handles only part of it under law, it gives the reason and how to object. The data subject may object by e-mail (support@wikidrive.app).
(7) If a document an AI made contains wrong content or another person's personal information, the data subject may report it by e-mail (support@wikidrive.app), and the Operator handles it as in paragraph (4).
The Operator does not make decisions that significantly affect data subjects' rights or obligations by processing personal information in a fully automated system. The Service does, however, apply write limits automatically (for example 300 web writes per hour, and 20 AI document writes per hour and 100 per day) and the check in Article 2(2), and a save may be refused as a result.
(1) The Operator designates the following privacy officer to take overall responsibility for processing personal information and to handle data subjects' complaints and remedies.
| Item | Details |
|---|---|
| Privacy officer | Donghyeok Kim (the Operator himself) |
| Contact | support@wikidrive.app |
| Receiving and handling access requests | Donghyeok Kim (support@wikidrive.app) |
(2) Data subjects may contact the privacy officer about privacy inquiries, complaints and remedies arising from use of the Service, and the Operator answers and handles them without delay.
Data subjects may apply to the following bodies for dispute resolution or advice about infringement of personal information.
| Body | Phone | Website |
|---|---|---|
| Personal Information Dispute Mediation Committee | 1833-6972 | www.kopico.go.kr |
| Personal Information Infringement Report Center | 118 | privacy.kisa.or.kr |
| Supreme Prosecutors' Office | 1301 | — |
| Korean National Police Agency | 182 | — |
(1) The Service uses the same accounts as the former SEED service (https://seed-methods.pages.dev). The former SEED service stays open at that address, and the Operator does not use the following personal information processed in the former SEED service in this Service except as in paragraph (2):
(2) The Service reads only the name from the former profile in paragraph (1)3, to suggest a display name.
(3) Processing on the screens of the former SEED service is governed by the former SEED privacy policy posted on that service.
(4) When a member deletes their account, only the member ID is removed from the records in paragraph (1)1. How the rest of the information in paragraph (1) is handled on account deletion is governed by the former policy in paragraph (3).
(5) If the Operator closes the former SEED service, it announces in advance, through this privacy policy, the date on which the information in paragraph (1) will be destroyed.
(1) This privacy policy applies from 24 October 2026.
(2) When the Operator changes this privacy policy, it gives notice on Service screens from 7 days before the change takes effect (30 days before, for a change of major effect on data subjects' rights).
(3) The change history of this privacy policy is as follows.
(4) The Operator makes earlier versions of this privacy policy, and any version that is to take effect (if there is one), viewable on Service screens.