Privacy Policy
English reference translation. This English version is provided for reference only. The Korean version is the binding version, and if the two versions differ, the Korean version prevails. This policy is effective from November 19, 2026; the provisions on paid payments (Section 1.A(8)) are effective from November 26, 2026.
Atelier Launch (the "Service") is operated by Atelierlaunch (representative: Jung Sung Hun, the "Operator"). In accordance with Article 30 of the Personal Information Protection Act (개인정보 보호법), the Operator establishes this Privacy Policy to protect users' personal information and to handle complaints promptly.
- Service: a web service that helps solo founders validate ideas, plan, start development and run their service (currently in beta; validation beyond the free tier is paid), https://atelierlaunch.kr
- Operator: Atelierlaunch (representative: Jung Sung Hun, business registration number 367-35-01904, 1st floor, 129-3 Myeongan-ro 18beon-gil, Dongnae-gu, Busan (Allak-dong), Republic of Korea), contact atlier.skill@gmail.com
- From November 26, 2026, validation beyond the free tier and the Virtual Residents "Fix Order" can be purchased one at a time on the checkout page of Groble (groble.im). Payment method information such as card numbers, and the buyer's name and contact details, are received by Groble; the Operator does not receive them (Section 1.A(8), Section 4).
1. Purposes of Processing and Items of Personal Information Processed
A. Personal information processed without consent (conclusion and performance of a contract)
The information below is essential to provide the service when a user signs up or directly uses a feature, so it is processed without consent under Article 15(1)4 of the Personal Information Protection Act (conclusion and performance of a contract, or measures requested by the data subject in the process of concluding a contract). The Operator bears the burden of proving that the information is essential (Article 22(3) of the same Act).
(1) Sign-up and login
| Item | Purpose | How collected |
|---|---|---|
| Email address (ID), sign-up time | Identifying members, login, responding to inquiries | Sign-up screen, Google or Kakao account login |
| Google account link information (account number assigned by Google, time linked) | Login with a Google account | Google account login (if the user chooses it) |
| Kakao account link information (member number assigned by Kakao, time linked) | Login with a Kakao account | Kakao account login (if the user chooses it, from November 4, 2026) |
| Password (only an irreversibly transformed value) | Login | Sign-up, password recovery |
| Recovery code (transformed value only), time created, time last used | Recovering access when the password is forgotten | Sign-up, account screen, password recovery |
| Login records (session number, time created, expiry time) | Keeping you logged in (30 days) | Automatically at login |
| Usage status: time the free full version was used, number of quick checks used, credit grant and usage records, time of account deletion request or suspension | Managing free limits and credits, processing account deletion, responding to abuse | Automatically during use |
- We do not collect names separately. The name field in member information is automatically filled with the part of the email address before the @.
- Passwords and recovery codes are not stored in their original form; only irreversibly transformed (hashed) values are stored. Even the Operator cannot know your password.
- Your email address is used as your ID, and when you sign up we send one confirmation email to that address (link valid for 24 hours). Clicking the confirmation link logs you in right away in that browser, and you can keep using the service even without confirming. If you forget your password, you can receive a reset email at the address you signed up with (link valid for 15 minutes), and setting a new password through that link also counts as completing email confirmation. Do not sign up with someone else's email address. If you cannot recover access by email, use the recovery code you created on the account screen. If you have also lost your recovery code, please contact us. We will create a new code only when we can confirm that you are the account holder.
- Login records (sessions) do not store IP addresses or browser information.
- If you choose Google login, we receive your email address from Google. We do not store your name, photo or Google access permission (token). Accounts signed up only through Google have no password or recovery code.
- If you choose Kakao login, we receive only your Kakao Account email address (optional consent) from Kakao. We do not receive or store your nickname, profile photo or Kakao access permission (token). If you do not consent to providing your email address, you cannot sign up or log in with Kakao, and we guide you to sign up with email. Accounts signed up only through Kakao also have no password or recovery code.
- If you log in with a Google or Kakao account that has the same email address as an account already registered with a password, the two accounts are not linked automatically. Because email confirmation is not required at sign-up, this prevents you from being connected to an account that someone else created first.
(2) Idea validation and one-page plan (viewable only by you)
| Item | Purpose | How collected |
|---|---|---|
| One-line idea (up to 2,000 characters) or area of interest, who it is for, why you want to build it, project title (only the idea is required; the rest only if you want to enter them) | Creating AI validation results (problem statement, competitor research, assumptions, business viability, stop criteria, etc.), your project list | New idea screen |
| "Things to fix" added at each step (up to 500 characters), the idea chosen from candidates suggested by the AI | Reflecting them in the next step's results | Validation screen (only if you want to enter them) |
| Step results generated by the AI, validation report, AI response summary (URLs of web sources consulted, usage) | Showing, downloading and revisiting results, handing off to development | Automatically during validation |
| One-page plan (summary, features, screens, tasks for people, costs), the first draft written by the AI, number of revisions | Creating, revising and finalizing the plan, putting it into the development repository | Planning screen |
| Requests for the plan draft (up to 300 characters). Not stored; used only in the AI request | Reflecting them in the draft | Planning screen (only if you want to enter them) |
| Answers to the plan questions (who will use it, the moment they look for it, how they do it now, narrowed target, must-have items, how it makes money, form, deadline, cost, assumptions to check first, requests; if the verdict is "stop", what you want to keep and the changed idea). Personal information patterns are masked and the answers are encrypted when stored | Tailoring the plan to your situation, continuing to answer later | Questions on the planning screen (only if you want to answer; all can be skipped) |
| Experiment records: experiment conditions (hours available per week, whether you can post in communities, names of communities you already use), numbers entered for each task (number of posts and comments found, visits, sign-ups, etc.), sign-up page URL, notes. Community names, URLs and notes are encrypted and visible only to you | Choosing the next experiment task on the task screen, showing progress for each assumption | Task screen (only when entered) |
| Sign-up page: service name, one-sentence description of what it is, description, pain question and three pain lines, three feature lines, price, target, social media URLs for launch news (Instagram, Threads, X, YouTube, Naver Blog and Kakao Open Chat only), theme. Stored encrypted; once published, anyone who knows the page URL can see it | Publishing the sign-up page the user created and showing the counts of visits, interest and answers | Task screen, validation screen (only when you create one) |
| If you save a Virtual Residents result: the service URL that was read, the map read from the public page at that URL (content and position of text, buttons, links and input fields), result numbers (average star rating, number of residents who clicked through to the end, number of residents who left on the first screen), time and method of saving. Screenshots are not stored | Revisiting past results in your project's Residents tab, comparing how the numbers changed after fixes | "Save this result" on the Virtual Residents result screen, "Send them in again" on the Residents tab (only when you click) |
| If you keep My Resident: answers chosen for 10 questions (age group, type of work, interests, habits when looking at a service), the name you gave it | Having My Resident browse along every time you enter a URL, showing My Resident on the residents' home screen | "Keep My Resident" on the Virtual Residents result screen (only when you click) |
| My Resident's browsing records (if you keep My Resident): URLs of services that other users entered publicly into Virtual Residents and that match its interests, the stars My Resident gave, whether it clicked through to the end, where it left and why, device type, time spent, what it said while browsing and its answers to questions (text generated by the engine). Screenshots and maps are not stored | Viewing recently browsed places on My Resident page | Automatically once you keep My Resident (once an hour, and when you open My Resident page) |
| If you receive My Resident news emails (turned on by default when you keep My Resident): the email address you signed up with, when emails were sent and whether sending succeeded | Sending a summary of places My Resident browsed every Monday morning (not sent in weeks with no browsing), notifying you right away when it clicks through a place to the end (up to one email per day) | Automatically when you keep My Resident. You can turn it off anytime with "Stop receiving emails" in the email or the switch on My Resident page |
- Ideas, reports and plans may be trade secrets, so only you can view them. In admin tools and statistics, the Operator does not see their content and sees only numbers (number of runs, costs, etc.).
- Sign-up page visitors: we do not collect personal information such as names or email addresses. "I'm interested" and the chosen options (current method, how often, at this price) are added up only as counts by date, and who chose what is not recorded. To count the same person only once a day, a value transformed from the connecting IP in a way that changes daily (without the original IP) is kept for two days. Visits by the page's creator are not counted.
- The question conversation itself is not sent to the AI. Your answers to the questions are sent along with the plan request only when you click "Create a plan with this".
- If your input contains patterns of phone numbers, resident registration numbers, email addresses, account numbers, card numbers, apartment building and unit numbers, or names, they are replaced with a "[masked]" marker before being stored or sent to the AI. Some information cannot be detected by pattern, so do not enter sensitive information, unique identifiers such as resident registration numbers, or other people's personal information.
(3) Quick check before sign-up (only when you click start on the home screen, from October 7, 2026)
| Item | Purpose | How collected |
|---|---|---|
| One-line idea (up to 2,000 characters) | Creating the pre-sign-up quick check result (idea summary, problem statement, similar services), moving it to that account's projects upon sign-up | Home screen |
| Quick check result generated by the AI, AI response summary (URLs of web sources consulted, usage) | Showing part of the result, revealing all of it upon sign-up | Automatically during the quick check |
| Preview verification value (transformed value of a token kept only in the browser) | Showing the result only in the browser where it was started, moving it after sign-up | Automatically when the quick check starts |
- Ideas from people who have not signed up are not linked to identity information such as email addresses, and the result can be viewed again for 24 hours in the browser where it was started and in a browser that opens the link received by email (the number and verification value are kept only in that browser and can be deleted right away with "Delete from this device"). Personal information patterns are replaced with a "[masked]" marker before storage and before sending to the AI. If you ask to receive the pre-sign-up result link by email, we send one email with a link back to that result to the email address you entered. The address is used only for sending and is not stored; only a transformed value used to prevent repeated sending to the same address is kept for 24 hours. The email does not contain the idea or result content.
- If you prefer, you can sign up first with "Start for free" at the top of the home screen.
(4) Handing off to development (only if you want)
| Item | Purpose | How collected |
|---|---|---|
| GitHub access token (scope "repo", encrypted, up to 1 hour) | Creating one new repository in the user's account and uploading the package files (validation report, plan, roadmap, instructions), attaching a progress-signal webhook to that repository (from October 15, 2026), building and downloading the APK of the store app | Granting permission on the GitHub screen after "Connect with GitHub" (also "Reconnect GitHub to turn on" in the Development tab, and "Connect GitHub" for the APK) |
| GitHub ID (login name) | Building the repository URL, showing the connected account | Received from GitHub with the token above |
| Repository name, visibility, default branch, commit number, number of files uploaded, progress step, time, error type, answers about prerequisites (paid Claude subscription, whether you have a GitHub account, etc.) | Continuing after closing and reopening the window, reopening the created repository, daily creation limit | Input on the handoff screen, automatic |
| Transformed value of the first-instruction URL (prompt_url), time first opened, "I opened it" (web/app) | Delivering the first instruction to Claude Code, checking progress | Automatic on the handoff screen |
| Progress-signal webhook number, signing secret (encrypted), progress records for each checkpoint (checkpoint, status, the one-line summary Claude wrote in the commit message, preview URL, screenshot path (encrypted), commit number, time received) (from October 15, 2026) | Showing development progress in the Development tab | Received from GitHub through the repository webhook (not received if no webhook is attached) |
| "Email me when a check is waiting" on/off (from October 15, 2026) | Checkpoint notification emails | Development tab (only when turned on) |
| Sign-up for "Email me a check-in in 3 days", and the scheduled and actual send times | Sending one email reminding you to check the "To-dos this week" in the validation report | Report screen (only if you sign up; off by default) |
- GitHub tokens are encrypted (AES-GCM) and kept for up to 1 hour, and are used only for the following.
- Handoff: creating a new repository, uploading files to that repository, attaching one progress-signal webhook to that repository. Once the repository and webhook are created, the token is deleted immediately and also revoked (invalidated) on GitHub. When only the webhook is reattached from the Development tab, the token is also revoked right after attaching it.
- Building and downloading APKs (store app, only when connecting from the Development tab): running that repository's APK workflow, viewing recent run records and the list of result files, downloading the result file (zip). This token is kept and used for 1 hour, and files in the repository are not modified. APKs are not stored on the server; each time you download, the APK is fetched from GitHub and passed on as is. The QR URL for downloading on a phone is open for only 10 minutes, and during that time anyone with the URL can download without logging in, so do not send it to others.
- Tokens older than 1 hour (unfinished handoffs, APK connections) are deleted and revoked by the cleanup job that runs every hour.
- Development progress records: once a webhook is attached, GitHub sends the full push notification (commit messages, changed file paths, the committer's name and email address, and other content determined by GitHub) every time you push to the repository. After verifying the signature, the server selects and stores only the lines in commit messages that start with
Atelier-Progress:, the paths of screenshot files indocs/progress/, and commit numbers, and discards the rest without storing it. Code and file contents are not included in push notifications, so we do not receive them. Summaries, URLs and paths are encrypted and visible only to you. You can delete the webhook anytime in the repository settings, and once it is deleted we receive nothing more. If you delete your account or the project is permanently deleted, the signing secret is also deleted and notifications arriving after that are not stored, but the webhook itself remains in the repository, so please delete it in the repository settings. - The first-instruction URL (prompt_url) is a random value that cannot be guessed; the server stores only its transformed value, and the URL stops opening after 7 days. This URL contains only the instruction text, with no email address, repository name or token.
- What is sent to the repository is described in Section 3 (Provision to Third Parties).
(5) Operations monitoring and managed service (only if you want)
| Item | Purpose | How collected |
|---|---|---|
| Service URL entered by the user (only public URLs starting with https) | Opening it once an hour (every 5 minutes if the managed service is on) to check whether it is up and how fast it responds | Operations screen |
Check results (success or failure, response code, response time, error type, time); for the managed service, also whether /health is OK and its version value; checks on this week's to-dos | Showing uptime, recent records and to-dos | Automatic checks, operations screen |
| Notification method (email on/off) and member email address | Sending emails about service outages and recovery, and notices about low credits or suspension of the managed service | Email notifications turn on when the managed service is turned on (can be turned off on the operations screen) |
| Operations summary: the key flow extracted by code from the finalized plan and validation report (up to 5 lines), key assumptions (up to 3), names of key steps. Decision criteria (sign-up rate threshold, number of visits needed for a decision) | Basis for operations reports, step names in the connection snippet, numeric verdicts | Automatic. Decision criteria on the operations screen |
| Daily, weekly and monthly operations reports (numbers for that period, observations and suggestions written by the AI), reactions to reports (👍/👎) | Interpreting operations numbers, suggesting what to try next | Managed service, automatically every day (AI) |
| Daily to-dos (title, reason, text to paste into Claude Code), backup candidates, done / reason for removal (preset options) | Showing today's to-dos | Same as above, operations screen |
| Choice of promotion channels, promotional post drafts (AI), date posted | Drafts of posts for the chosen channels | Only when the managed service's promotion feature is turned on, automatically after the daily report (AI) |
- We receive only the URL. We do not receive secrets, tokens or admin accounts of the user's service. We do not read the response body of registered URLs. When the managed service is on, we read the
/healthresponse body at the same URL only up to 64KB, keep only whether it is OK and its version value, and discard the rest. - When creating operations reports, to-dos and promotion drafts, the original idea text and the full plan are not sent to the AI. The items sent are listed in Section 5.A.
(6) Store app operations (only if you want, from October 16, 2026)
| Item | Purpose | How collected |
|---|---|---|
| Minimum and latest versions, Google Play and App Store URLs | Update notices in the app (the app reads them from a public URL) | Operations screen |
| Store numbers copied in by the user (new installs per week, average star rating, number of ratings, crash rate) | Showing the change from the previous week | Operations screen |
| Store description draft (AI) | Helping write the store listing text | Operations screen, AI (Anthropic, Section 4) |
| Pasted review text | Creating reply drafts | Operations screen, AI. Not stored (phone number and email patterns are masked before being sent to the AI); only the daily count is recorded |
(7) Sending feedback (only if you want)
| Item | Purpose | How collected |
|---|---|---|
| Feedback written by the user (up to 500 characters), feedback type (inconvenience, strange result, feature suggestion, other), type of screen it was sent from (home screen, validation, planning, etc.), member number (only if logged in), time sent | Checking inconveniences, improving the service | The round "Send feedback" button at the bottom right of every screen (can be sent without signing up) |
- If feedback contains personal information patterns such as phone numbers, email addresses or resident registration numbers, they are replaced with a "[masked]" marker before storage. Only the Operator sees it.
(8) Paid payments: validation and Fix Order (only when you pay, from November 26, 2026)
| Item | Purpose | How collected |
|---|---|---|
| Payment number (a random number generated by the Operator), product purchased (quick check, full validation, continue to full version, Fix Order), price and actual amount paid, Groble payment identifier, payment status (before payment, paid, used, refunded), payment, use and refund times, member number. For validation, the project number and the number of the validation started with that payment; for Fix Order, the Virtual Residents result number and the numbers of the selected fix candidates | Confirming payment, starting the purchased validation and using Fix Order, processing refunds, handling disputes | The payment number is created when you click "Pay"; the amount paid, identifier and times are received through a notification signed by Groble when payment is completed |
- This applies from November 26, 2026. Payment takes place on the checkout page of Groble (groble.im), and the buyer's name, contact details and payment method information are received directly by Groble under its privacy policy. Even if buyer information comes along in Groble's notification, the Operator neither reads nor stores it.
- Even if you delete your account or a project, or the Virtual Residents reading record is deleted after 7 days, payment records are kept separately for the period set out in Section 2. The text of a purchased Fix Order (names of the things to fix, how to fix them, instructions to give Claude) is attached to the saved Virtual Residents result and is kept and deleted together with that result. While Fix Order is on sale, when Virtual Residents read a URL, the text of each section of that public page, up to 1,500 characters per section, is stored together in the reading record (kept for 7 days, the same as the reading record) and is used only when using Fix Order.
B. Personal information processed without consent (service protection and statistics)
The information below is needed to protect the service from abuse and to fix errors. It is processed under Article 15(1)6 of the Personal Information Protection Act (legitimate interests of the personal information controller), as a case where this interest clearly overrides the user's rights and the processing does not go beyond a reasonable scope.
| Item | Purpose |
|---|---|
| Value of the connecting IP transformed (hashed) in a way that changes daily. When signing up for launch notices, starting a free validation, entering a recovery code, sending feedback before sign-up, or clicking a site or opening a share link counted for the Virtual Residents home screen board | Preventing abuse (launch notice sign-ups 20 per day, free full version 2 per day from the same connection, free quick check 6 per day, beta full version 3 per day, recovery codes 5 per 15 minutes, feedback before sign-up 5 per day, board reactions counted once a day per person) |
| Transformed value of the member number. When the password is entered incorrectly while creating a new recovery code | Preventing abuse (5 per 15 minutes) |
| Transformed (hashed) value of the email address. When sending email confirmation or password reset emails | Preventing email abuse (5 emails per 24 hours to the same email address) |
| Connecting IP and request path. For account requests such as login and sign-up | Preventing password guessing and sign-up abuse (login 10 per 15 minutes, sign-up 10 per day) |
| Value of the connecting IP transformed in a way that changes daily, temporary page number, referral source. When visiting the home screen. When opening the Virtual Residents screen, the temporary page number and referral source (only the domain of the referring site, from November 18, 2026), and whether this browser has opened it before and whether it has run a check (yes or no) | Estimating visit counts |
| What the visit analytics tool (PostHog) collects: URLs of screens viewed (with the part after the question mark and share numbers removed), where you came from, time spent, where you clicked and scroll position, screen size and browser type, screen recordings (with all text and input masked), a random number that distinguishes this browser, and the names and option codes of screen events among the usage records above. From November 6, 2026 | Improving the service by seeing which screens people stay on and which they leave (Section 10) |
| Usage records (below). Email addresses, idea content, titles, service URLs, repository names and GitHub IDs are not included; for members, only the member number | Service improvement, cost management statistics |
| Service access and error records (Cloudflare processing logs). Passwords and idea content are not recorded | Checking errors, security |
- Usage records: home screen visits, visits that actually viewed the home screen (time spent), answers to check questions (option codes), whether you signed up and whether you chose an option on willingness to pay, temporary page number / sign-up, login, password recovery, account deletion (days since sign-up, number of projects) / project creation (whether an idea was entered, whether anything was masked), validation start, step completion, failure (type, step, model, token count, cost, number of searches, time taken, error type), report download, reaching a limit / plan creation, failure, finalization (number of items, whether revised), progress on plan questions (question name, number, whether skipped; no answer content) / connecting or disconnecting an operations URL (without the URL), to-do checks, experiment conditions (hours, whether you post, number of communities), experiment task handling (track, step, numbers entered) / handoff progress (step, time taken, prerequisite answers), GitHub connection, repository creation (visibility, number of files), failure type, first instruction opened, "I opened it" (web/app) / whether attaching the progress-signal webhook succeeded, reaching a checkpoint (checkpoint, status), copying the things-to-fix text (checkpoint) / APK build start, APK download / managed service start, deduction, pause and cancellation, notification sending (type, channel, success), creating operations reports, promotion drafts and store drafts (type, cost, check result), report reactions, to-do handling (option code for the removal reason), time.
- The Operator does not store original IP addresses in the database and keeps only transformed values. However, the records of login and sign-up request counts contain the IP as is, and they are deleted 1 day after the last request (Cloudflare processing logs are also an exception).
- The temporary page number is a random value newly created each time a screen is opened; it is not stored in a cookie and does not link the same person across different days. (The visit analytics tool's browser ID counts the same browser as the same across different days: see Sections 9 and 10.)
- Public posts found during validation: for public community posts whose originals the code opened and checked during competitor research, the URL, a quote (under 120 characters, with personal information patterns masked), post type and year written are separated from the idea, kept in a "pain post collection" for 13 months, and used for service improvement and as evidence for other validations. This collection does not keep ideas, member numbers or search terms.
C. Personal information processed with consent
| Category | Items | Purpose | Legal basis |
|---|---|---|---|
| Launch notice sign-up (waitlist) | Email address, time of consent (required). Willingness to pay (whether you would like to buy credits (1 credit is about 1,000 won) after launch and continue using the service through development) (optional). Answers to check questions (where you got stuck, development experience, whether you have a paid Claude subscription, whether you have a GitHub account, as preset option codes), referral source (utm) (if you answered the check) | Sending service launch notices, deciding features and pricing, service improvement statistics | Consent of the data subject (Article 15(1)1) |
| Development tab early access sign-up | Email address, time of consent (required). Whether you are currently building a service ("Yes" or "Just starting") (optional). Referral source (utm) | Sending development tab early access invitation emails, trial guidance and feedback requests, and free pass notice emails | Consent of the data subject (Article 15(1)1) |
| "Make My Resident" opening notice sign-up | Email address, time of consent (required). Answer chosen in the window ("Notify me when it opens" or "I'm interested"), referral source (utm) | Sending one email announcing that "Make My Resident" has opened | Consent of the data subject (Article 15(1)1) |
| GitHub repository creation | Table in Section 3 | Creating a repository in the user's own GitHub account | Consent of the data subject (Article 17(1)1, Article 28-8(1)1) |
- You can still sign up and use other features without consenting. If you do not sign up for launch notices, early access or "Make My Resident" notices, you will not receive notice, invitation or notification emails; if you do not consent to GitHub repository creation, you can download the report as Markdown and use it yourself.
D. Information not processed
We do not receive real names, phone numbers, unique identifiers such as resident registration numbers, sensitive information, payment information, location information, GitHub passwords or secrets of the user's service. Therefore there is no possibility of sensitive information being disclosed. We also do not process pseudonymized information.
2. Processing and Retention Periods of Personal Information
| Information | Retention period |
|---|---|
| Member information (email address, sign-up time, password and recovery code (transformed values), usage status, credit records) | Until account deletion. When you delete your account, use stops immediately and you are logged out, and shared URLs and sign-up pages are also closed right away. For 6 days you can log in again to undo the deletion; after that, the data is permanently deleted by the cleanup job (within 7 days at the latest) |
| Login records (sessions) | 30 days (extended once a day while in use). Ended on logout, password recovery or account deletion |
| Pre-sign-up quick check (idea, result, preview verification value) | 24 hours. If you do not sign up, the hourly cleanup job deletes it (within 25 hours at the latest). If you sign up, it is moved to that account's projects and follows the row below |
| Ideas, step results, reports, plans, plan question answers, experiment records, sign-up pages (content and counts by date), development progress records, saved Virtual Residents results | Until the member deletes them. When you delete a project, you can restore it for 30 days, after which it is permanently deleted. If you delete your account, deleted within 7 days |
| Kept My Resident (answers chosen for the questions, name) | Until the member deletes their account. If you delete your account, deleted within 7 days |
| My Resident's browsing records | Until the member deletes their account (up to 200 records per person; beyond that, the oldest are deleted first). If you delete your account, deleted within 7 days |
| IP transformed value for counting sign-up page visits once a day | Two days. Deleted by the daily cleanup job (within three days at the latest) |
| AI response summary (URLs of web sources consulted, usage) | Cleared 30 days after the step ends (the report remains) |
| Plan creation count records (for the daily limit) | 2 days |
| GitHub access token (encrypted) | Up to 1 hour. Deleted immediately and revoked on GitHub once the repository and webhook are created. Connections for building APKs and unfinished connections are deleted and revoked by the hourly cleanup job after 1 hour. Deleted immediately on account deletion |
| GitHub connection verification value (state, transformed value) | Cannot be used after 10 minutes; deleted by the hourly cleanup job |
| First-instruction URL (prompt_url, transformed value) | Stops opening after 7 days and is deleted in the next cleanup job. Deleted immediately on account deletion |
| Handoff records (GitHub ID, repository name, visibility, commit number, step) | Same as the project. Deleted 30 days after the project is deleted, or within 7 days of account deletion (the repository itself on GitHub belongs to the user and is not deleted) |
| Repository creation count records (for the daily limit, without repository names) | 2 days |
Operations monitoring URL, last check result (including the /health version value), notification method | Until disconnected (check records are also deleted immediately on disconnection). Deleted 30 days after the project is deleted, or within 7 days of account deletion |
| This week's to-do checks | Same as the project. Deleted 30 days after the project is deleted, or within 7 days of account deletion |
| Operations summary, decision criteria | Same as the project (kept even after disconnection). Deleted 30 days after the project is deleted, or within 7 days of account deletion |
| Operations reports (numbers, AI text, reactions), choice of promotion channels and promotional post drafts | Until disconnected (no separately set period; deleted immediately on disconnection). Deleted 30 days after the project is deleted, or within 7 days of account deletion |
| Sign-up for the check-in email in 3 days | Deleted 30 days after the scheduled send time (deleted immediately if turned off before sending). Also deleted 30 days after the project is deleted, or within 7 days of account deletion |
| Daily to-dos | To-dos completed, removed or replaced: 90 days from that day; unused backup candidates: 14 days from the day they were created. To-dos still open: until disconnected. Deleted immediately on disconnection, 30 days after the project is deleted, or within 7 days of account deletion |
| Individual check results (failed checks, results of checks run manually) | 7 days. Deleted immediately on disconnection |
| Outage records (status changes), notification records, daily uptime numbers | 13 months. Deleted immediately on disconnection |
| Store numbers, description drafts, version settings | Until disconnected (deleted immediately on disconnection). Deleted 30 days after the project is deleted, or within 7 days of account deletion |
| Review reply draft count records (without text) | 2 days |
| Feedback content | 1 year. On account deletion, the member number is deleted so that the sender cannot be identified |
| Waitlist information (email address, time of consent, willingness to pay, check answers, referral source) | 6 months after the launch notice is sent. If the notice has not yet been sent, 1 year from the sign-up date. Without delay if you click the "Remove me from the list" link in the launch notice email or request withdrawal |
| Early access sign-up information (email address, time of consent, whether you have a service, referral source) | 6 months from the sign-up date. Without delay if you request withdrawal |
| "Make My Resident" opening notice sign-up information (email address, time of consent, chosen answer, referral source) | 1 year from the sign-up date. Once the notice is sent, 30 days after that. Without delay if you request withdrawal |
| Domains on the Virtual Residents village board (without the path or anything after the question mark) | 30 days from the last day it was entered publicly. Daily reaction counts used for the board ranking (numbers only) are kept for 8 days |
| Abuse-prevention transformed values (IP, member number, email address) | 24 hours |
| Login and sign-up request count records (including IP) | 1 day after the last request |
| Visit count estimation hashes | Deleted after being aggregated into the next day's visit count (a number) (around 24 hours) |
| Daily usage signal numbers (Section 11) | 13 months. Deleted immediately on disconnection |
| Daily transformed values of web visits and app install numbers (Section 11) | Deleted after being aggregated into numbers the next day (around 24 hours) |
| Usage records (statistics) | 3 years from the date recorded. On account deletion, the member number is deleted so that it cannot be identified whose record it is |
| Operator admin records (credit grants and withdrawals, suspensions, recovery code reissues; member number and type of action only) | 3 years (numbers of members who deleted their accounts are not linked to other information) |
| Paid payment records: validation and Fix Order (Section 1.A(8)) | 5 years (Article 6 of the Enforcement Decree of the Act on the Consumer Protection in Electronic Commerce (전자상거래 등에서의 소비자보호에 관한 법률)). Kept separately for this period even if you delete your account or project, and deleted by the cleanup job after the period ends |
| Processing logs (Cloudflare) | 7 days |
| Backups (Cloudflare D1 automatic restore points) | Deleted information remains in automatic backups for up to 30 days and then disappears |
| Retention by processors (Anthropic, email delivery provider) | Section 5 |
The information preserved separately under law is the paid payment records (validation and Fix Order). They are preserved under Article 6 of the Act on the Consumer Protection in Electronic Commerce and Article 6 of its Enforcement Decree (records on payment and on the supply of goods, etc., 5 years), and during that period they are used only to confirm payments, process refunds, handle disputes and comply with legal obligations.
3. Provision of Personal Information to Third Parties
The Operator processes personal information only within the purposes stated in Section 1 and does not provide it to third parties except in the following cases.
- When the user directly requests it and consents (table below)
- When there is a special provision in law, or when it is unavoidable to comply with a legal obligation
The entrustment of processing in Section 4 means work is assigned on the Operator's behalf, and the recipients do not use the information for their own purposes, so it is not provision to a third party.
Provision at the user's request: GitHub repository creation (handing off to development)
| Recipient | Purpose | Items | Recipient's retention period |
|---|---|---|---|
| GitHub, Inc. (United States). The user's own GitHub account | Creating the new repository requested by the user and uploading the package files | Project title, one-line idea, target, validation report, finalized plan, roadmap, instruction files. Repository name and visibility (private by default) | Until the user deletes the repository on GitHub (subject to GitHub's terms and privacy policy) |
- Timing and method: sent over an encrypted network (GitHub API) when the user clicks "Connect with GitHub", grants permission on the GitHub screen, and then clicks "Create repository".
- This is an overseas transfer, so please also see Section 5.B.
- You have the right to refuse consent. If you refuse, you cannot use the handoff to development, but you can download the report as Markdown and use it yourself, and other features are not affected.
- The repository created belongs to the user. The Operator does not delete or modify the repository. Who can see the contents of the repository depends on the visibility the user chose.
4. Entrustment of Personal Information Processing
The Operator entrusts the following tasks to others in order to provide the service. Processors use personal information only for the entrusted tasks and not for their own purposes. Personal information protection obligations are set out in the entrustment contracts (each company's data processing terms).
| Processor | Entrusted tasks |
|---|---|
| Cloudflare, Inc. | Operating the web service (hosting), database storage, processing logs and backups, relaying AI requests, name resolution (DNS lookup) for operations monitoring URLs |
| Anthropic, PBC | AI processing (Claude API) that creates idea validation results, plan drafts, operations reports and to-dos, promotional post drafts, store description and review reply drafts, and Virtual Residents Fix Orders; web search for competitor research |
| PostHog, Inc. | Visit analytics: which screens people stay on and leave, screen recordings (text and input are masked). Email addresses, member numbers, ideas and IPs are not sent |
| Waffle (waffleweb.co) | Ad display and visit analytics on the Virtual Residents screen (Section 10). Service URLs, result content, email addresses and member numbers are not sent |
| Plus Five Five, Inc. (Resend) | Sending email confirmation emails, password reset emails (when requested), checkpoint notification emails, report check-in emails after 3 days (once, if you signed up), pre-sign-up result link emails (when requested), operations outage and recovery notification emails, managed service low-credit and suspension notice emails, and My Resident news emails (if you receive them) |
- Payments are taken on the checkout page of Groble (groble.im) (from November 26, 2026). The name, contact details and payment method information that buyers enter at Groble are processed by Groble under its own privacy policy, and the Operator receives only the amount paid, the payment identifier, and the payment and refund times from Groble.
- If processors change, we will revise this policy and give notice.
5. Overseas Transfer of Personal Information
A. Overseas transfer for entrusted processing and storage
To conclude and perform the contract for providing the service, personal information is entrusted for processing and stored overseas as follows (Article 28-8(1)3 of the Personal Information Protection Act, disclosed in this policy).
| Recipient (contact) | Country of transfer | Items transferred | When and how transferred | Purpose of use | Retention period |
|---|---|---|---|---|---|
| Cloudflare, Inc. (sar@cloudflare.com) | US corporation. Requests are processed at the Cloudflare data center near the user, and the database is stored in a data center in a region determined by Cloudflare. AI requests are passed on to Anthropic through a data center in the eastern United States (relayed only, not stored) | All items in Section 1 | Sent over an encrypted network each time the service is used | Hosting, storage, logs, backups, relaying AI requests | Same as Section 2 |
| Anthropic, PBC (privacy@anthropic.com) | United States | Idea (or area of interest), target, reason, chosen candidate, things to fix, requests for the plan and answers to the plan questions (after personal information patterns are masked, sent together when the plan is created), previous step results, validation report, idea entered in the pre-sign-up quick check. Managed service operations reports and to-dos: usage numbers by period (visits, page views, sign-ups, logins, number of payments and total amount, numbers of feedback items, errors and key steps, counts by top referring places (domains), tracking tags and most-viewed paths, number of app opens, uptime and number of outages, store numbers entered by the user), operations summary (key flow, assumptions, step names), one line from the previous report and its reaction, titles and reasons of removed to-dos. Promotional post drafts: numbers for the past 7 days (visits, sign-ups, payments, number of feedback items), first 3 lines of the operations summary, recent drafts. Store description drafts: project title, plan summary (or the one-line idea if there is none), target, feature list (up to 6), search keyword candidates and names of other apps from the validation report, name and subtitle of recent drafts (after personal information patterns are masked). Review reply drafts: pasted review text (after personal information patterns are masked). Virtual Residents Fix Order (if purchased): the text of each section of that service's public page (up to 1,000 characters per section, after personal information patterns are masked) and the names of the chosen fixes. Email addresses, passwords and member numbers are not sent | Each time a validation step or plan draft is created (once validation starts, the remaining steps run in order, and the server continues even if you close the screen; when you click "Next" or "Create"; for the quick check, when you click start on the home screen; for the first text of a sign-up page, once per project when it is first created, with the validation report sent). Operations reports and promotion drafts are created automatically by the server every day after 9 a.m. (Korea time) while the managed service is on; store description and review reply drafts when the user clicks create. Fix Order once, when payment is confirmed. All sent over an encrypted network (HTTPS, Anthropic API) | Generating AI results, web search for competitor research | Deleted within 30 days of receipt (up to 2 years if flagged as a policy violation) |
| Plus Five Five, Inc. (Resend, support@resend.com) | United States | Recipient email address, email subject and body (email confirmation link, password reset link, checkpoint name and link, report to-do check-in notice and link (idea and to-do content are not included), pre-sign-up result link (idea and result content are not included), outage and recovery notification content (service URL, outage type, possible causes, diagnostic text), low-credit and suspension notices, My Resident news (My Resident's name, age group and occupation, URLs of services browsed, stars, where it left and why, what it said)) | Sent over an encrypted network each time an email is sent | Sending emails | The retention period of sending records follows Resend's privacy policy |
| PostHog, Inc. (privacy@posthog.com) | Germany (PostHog EU Cloud, Frankfurt) | Visit analytics items in Section 1.B. Email addresses, member numbers, idea content and IP addresses are not sent | Sent directly by the user's browser over an encrypted network each time a screen is viewed | Visit analytics (service improvement) | Screen recordings 30 days; other visit records follow PostHog's retention policy. If the Operator stops using the tool, the project is deleted and the data is deleted with it |
How to refuse and consequences
- If you do not want your information transferred overseas, do not sign up as a member or for launch notices. In this case you cannot use the service (the introduction screen and check questions can be viewed without signing up).
- For people who have not signed up, nothing is transferred to Anthropic unless they start a quick check on the home screen.
- Even after signing up, if you only want to avoid transfer to Anthropic, do not use idea validation or plan drafts. In this case you cannot receive validation results or plans, and you also cannot use the handoff to development or operations monitoring, which require a validation report. Operations monitoring sends nothing to Anthropic unless the managed service is turned on.
- If you do not want transfer to the email delivery provider, do not turn on checkpoint notification emails, do not sign up for "Email me a check-in in 3 days" in the report, and turn off email notifications for the managed service.
- If you do not want transfer to visit analytics (PostHog), turn on "Do Not Track" or "Global Privacy Control" in your browser. When it is on, nothing is sent from that browser. There is no disadvantage in using the service.
B. Overseas transfer based on user consent (GitHub)
GitHub repository creation in Section 3 sends information to GitHub, Inc. in the United States, so separate consent is obtained before the repository is created, under Article 28-8(1)1 of the Personal Information Protection Act.
- Recipient: GitHub, Inc. (privacy@github.com)
- Country of transfer: United States
- Items, purpose, retention period: same as the table in Section 3
- When and how: over an encrypted network (GitHub API) when you click "Create repository"
- How to refuse and consequences: if you do not consent, do not do the handoff; you can only download the report as Markdown and use it.
C. Information sent directly by the user's browser (external fonts)
All screens load the font (Pretendard) from jsDelivr (cdn.jsdelivr.net). At that time, the user's browser sends the connecting IP address, browser information and the address (domain) of the page visited directly to jsDelivr and its CDN providers. jsDelivr processes this information under its own policy, and the Operator does not receive it.
6. Procedures and Methods for Destroying Personal Information
- When the retention period ends or the purpose has been achieved, the cleanup job that runs every day at midnight (Korea time) deletes the data from the database. GitHub tokens and connection verification values are deleted by the cleanup job that runs every hour.
- Electronic files are deleted so that they cannot be recovered, and they also disappear from automatic backups after up to 30 days.
- No personal information is kept in paper documents.
- Paid payment records are preserved for 5 years as required by law and then deleted (Section 2).
7. Rights and Obligations of Data Subjects and Legal Representatives, and How to Exercise Them
- Users may at any time request access to, correction of, deletion of, or suspension of processing of their personal information, and may withdraw consent.
- Things you can do directly on screen: deleting ideas (projects) (can be undone within 30 days), revising plans, disconnecting operations monitoring URLs (reports, to-dos and drafts are also deleted), canceling the managed service, turning off notification emails, and "Delete account" on the account screen (deletes all data). Your waitlist entry can be deleted with the "Remove me from the list" link in the launch notice email.
- If you send a request to atlier.skill@gmail.com, we will process it within 10 days of receipt and tell you the result. Deletion of a suspended account can also be requested at this contact.
- If you believe someone else has entered your personal information into the service, please let us know at the same contact. Once confirmed, we will delete or mask it.
- Requests can also be made through a representative, such as a legal representative or a person you have authorized.
- Children under the age of 14 cannot sign up or sign up for launch notices.
8. Measures to Ensure the Security of Personal Information
- All communication is encrypted (HTTPS; requests coming in over http are redirected to https)
- Passwords are stored only with PBKDF2 (100,000 iterations), and recovery codes only as hashes. Sessions do not store IP addresses or browser information
- Ideas, reports and plans can be viewed only by their owner (checked on the server by member number). The Operator's admin tools do not show their content
- Personal information patterns are masked before storage and before sending to the AI. User input sent to the AI is separated as data, not instructions (defense against prompt injection)
- GitHub tokens are encrypted (AES-GCM) with a server secret and kept for up to 1 hour, and revoked right after the repository and webhook are created (for APK builds, after 1 hour). The progress-signal webhook accepts only notifications with a valid signature. The connection verification value (state) is checked by matching a browser cookie against a server hash, so that other sites cannot hijack the connection
- Operations monitoring opens only public https URLs and blocks internal networks, IP addresses and localhost
- Secrets such as Anthropic and GitHub app keys are kept only in the server's secret store. Access to the database and secrets is limited to the Operator alone, and the operator accounts (Google, GitHub, Cloudflare) use two-factor authentication
- Request rate limits (login 10 per 15 minutes, sign-up 10 per day, recovery codes 5 per 15 minutes, launch notice sign-ups 20 per day), daily cap on AI costs
- Connecting IPs are kept as transformed (hashed) values (except login and sign-up request count records, kept for 1 day). Passwords and idea content are not left in processing logs
9. Installation, Operation and Refusal of Devices that Automatically Collect Personal Information (Cookies, etc.)
- To keep you logged in, we use an essential cookie (session cookie, 30 days). If you block cookies in your browser, you cannot log in.
- While connecting to GitHub, we use an essential cookie for connection verification (10 minutes, deleted immediately when you return from GitHub).
- To count how many times this browser has used the quick check without sign-up, we use an essential cookie (
al_pv, 90 days). This cookie contains only the number of uses, the date of first use and a signature to prevent tampering, and does not contain names, email addresses, ideas or IPs. It is used only for the quick check without sign-up (the first 1 time); once you sign up, the member limit (3 times a month) applies. - To show the app's guide bubbles only once a day, the numbers and dates of guides viewed and the sign-up page visit and interest numbers are kept in this browser's storage (
al_guide). Idea content, names and email addresses are not included and are not sent to the server. They are deleted together when you clear browser site data, and the guides may appear again. - To reopen results run in Virtual Residents and compare them with the previous run, the service URL, result number, time run, star rating and a summary of residents' movements are kept in this browser's storage (
rs-history-v1) (the most recent 5 per service). Names and email addresses are not included and are not sent to the server. If you click "Save this result" on the result screen, that result is saved to a project in your account as described in Section 1.A(2). A resident made in "Make My Resident" (answers chosen for the questions and its name) is kept only in this browser's storage (atelier_twin) until you log in and is not sent to the server; once you log in through "Keep My Resident", it is saved to your account as described in Section 1.A(2). If you enter a URL into Virtual Residents publicly (the default), that service URL may appear on the My Resident pages of other users with matching interests. The domain of that service (without the path or anything after the question mark) is also posted on the village board on the Virtual Residents home screen, where anyone can see it, and this week's ranking is based on the number of clicks on the site name on the board and the number of times share links were opened (numbers only). Who entered it is not shown. If you turn on "Private" below the URL field before entering it, it will not appear. You can delete these with "Clear records in this browser" on the residents' home screen or by clearing browser site data. - To count whether the Virtual Residents screen has been opened before, only the date it was first opened is kept in this browser's storage (
rs-seen). Only whether the visit is a return visit and whether a check has been run (yes or no) are sent to the server. Names, email addresses and service URLs are not included, and the same person is not linked across different days. - For visit analytics (PostHog), a random number that distinguishes this browser and the visit count are kept in browser storage (items starting with
ph_). No cookies are used. Names, email addresses and ideas are not included. They are deleted together when you clear browser site data, and after that you are counted as a new visitor. - On the Virtual Residents screen, Waffle's code keeps whether this is a return visit and records of ad and notice display in this browser's storage (items starting with
waffle_orwf_). Waffle's code does not use cookies. Names, email addresses and service URLs are not included. Partner ads may use that ad company's cookies inside a separate window at Waffle's address. - How to refuse: you can block cookie storage in your browser settings (for example, in Chrome settings under Privacy and security, Cookies and other site data). If you block or delete cookies, this browser's usage count goes back to 0, so you can use the no-sign-up allowance again (abuse is separately prevented by a limit of 2 per day from the same IP, a service-wide daily limit and a trial cost limit).
10. Behavioral Information
From November 18, 2026, one ad slot from Waffle (waffleweb.co) is placed at the very bottom of the result screen on the Virtual Residents screen (atelierlaunch.kr/residents). The Waffle code on this screen sends Waffle the screen path (the result number after the question mark is not sent), the domain of the referring site, whether the device is a mobile phone, whether it is a return visit, time spent, and whether the ad was shown and clicked. Service URLs, result content, names, email addresses and member numbers are not sent. Waffle states that it uses the connecting IP and device information only as an anonymous key mixed with the date and does not store them (Waffle code documentation, checked November 18, 2026). When there is no Waffle ad, partner ads such as Kakao AdFit appear inside a separate window (iframe) at Waffle's address, and those ad companies may show personalized ads under their own policies. If you do not want ads, you can block them with your browser's ad-blocking feature, and there is no disadvantage in using the service. No ad tools are used on any other screen. The visit analytics tool for service improvement (PostHog, EU servers) has been used since November 6, 2026. The items collected are listed in Section 1.B, overseas transfer and how to refuse in Section 5, and browser storage in Section 9. All text and input are masked before sending, and information that identifies people (email address, member number) and IP addresses are not sent. Visit counts themselves are also counted on the server using the transformed values in Section 1.B.
11. Information the Operator Processes on Behalf of Users (usage signals of users' services, from October 16, 2026)
When a user turns on the managed service in the Operations tab and puts the connection snippet into their own service or app, the Operator receives usage signals from the people who use that service (the "end users") on behalf of the user and organizes them into numbers. The party responsible for this processing (the personal information controller) is the user who runs that service, and the Operator is the party entrusted with the processing (the processor). Users must describe this processing in their own service's privacy policy and in the store's privacy disclosures (the connection snippet explains this).
| Item | Purpose | How collected |
|---|---|---|
| Event name (visit, sign-up, login, payment, feedback, error, key step, app open, opened from notification), cleaned path (numbers and IDs removed), referring domain, tracking tags, payment amount (number only), time | Showing the user the numbers of visits, sign-ups and key steps | Signals sent by end users' browsers, apps and servers |
| Web: IP address (first 64 bits for IPv6), browser information, daily transformed value (hash) made with the date | Counting daily unique visits | Same as above. Original IPs are not stored |
| App: app version number | Share by version, update notices | Sent by the app |
| App: daily transformed value (hash) made from a random install number created when the app is first launched | Counting daily unique users | Sent by the app. Original install numbers are not stored. Not linked to advertising IDs, device IDs or accounts |
- We do not receive names, email addresses, contact details, input content or user numbers. Values outside the list are discarded. No cookies are used.
- The next day's cleanup job keeps only the number (that day's unique count) and deletes the daily transformed values. The transformed values cannot be used to link the same person across different days.
- The retention period is the same as the "Daily usage signal numbers" and "Daily transformed values of web visits and app install numbers" rows in the table in Section 2.
12. Privacy Officer and Complaint Handling
- Privacy officer: Jung Sung Hun (Operator)
- Contact: atlier.skill@gmail.com
- Requests for access, complaint handling and remedies for damage are also received at the same contact. We respond within 10 days of receipt.
13. Remedies for Infringement of Data Subjects' Rights
You can report personal information infringement and get counseling from the agencies below.
- Personal Information Dispute Mediation Committee (개인정보분쟁조정위원회): 1833-6972, www.kopico.go.kr
- Personal Information Infringement Report Center (개인정보침해신고센터): 118 (no area code), privacy.kisa.or.kr
- Supreme Prosecutors' Office (대검찰청): 1301 (no area code), www.spo.go.kr
- Korean National Police Agency (경찰청): 182 (no area code), ecrm.police.go.kr
14. Changes to the Privacy Policy
- When we change the content, we post a notice on the service screen from 7 days before the change takes effect (30 days before, if the change is unfavorable to users), and we also notify members and waitlist subscribers by email.
- Previous versions of the policy are kept separately so that you can view them: Previous Privacy Policy (before the November 19, 2026 revision) (Korean)
Effective date: This policy applies from November 19, 2026. The provisions on paid payments (Section 1.A(8)) apply from November 26, 2026.