Privacy Policy

Effective October 16, 2026

This Privacy Policy describes how Myworkingmemory LLC (the “Company,” “we,” “us,” or “our”) handles information for the separate Metadata Cleaner hosted service, its account connection, processing, downloads, and support (the “Services”). It is based on the existing native-app policy, with the different hosted data handling described below.

1. What this hosted service does

You select a Word .docx or PDF already available through ChatGPT. The service retrieves that selected file from its signed OpenAI download link and processes it on the Company’s existing Vercel-hosted Node service. This is an upload and server-processing service. It is not the native apps’ on-device processing.

Inputs are limited to 2 MiB per file. Cleaned outputs are limited to 3 MiB. The original is not modified. The service reports generic finding categories to ChatGPT, rather than raw author names, comments, filenames, or metadata values. It returns a temporary link to a separate cleaned copy when supported processing and verification finish.

Visible text, form answers, and metadata inside embedded images may remain. This is not redaction or anonymization. Unsupported, malformed, encrypted, or incompletely verified documents are refused.

The ChatGPT edition is for ordinary documents. Do not submit personal health records, payment-card or security details, government-ID numbers, passwords, API keys, or other authentication secrets. Before processing each file, you must acknowledge that input scope. This acknowledgement does not mean the service detects those contents or establishes regulatory compliance.

2. Documents, reports, and temporary downloads

The application downloads the selected original into memory for processing and does not write the original to application disk storage. Document contents and metadata are handled in memory to inspect the file and create the separate output. The application does not send documents to an AI model or use them for analytics or AI training.

Cleaned copies are stored as AES-256-GCM encrypted data in Neon. The temporary download capability contains the secret needed to decrypt its copy. Anyone who has that link can download the copy while the link remains valid. Keep the link private. The link expires after ten minutes and permits up to five fetches; it may stop earlier when access is disabled or the account grant expires.

Expired downloads become inaccessible immediately under the application’s access checks. Expired database rows are deleted during later processing or the existing scheduled daily cleanup when it runs successfully. This is not a promise that stored data is physically erased at the exact download-expiry time. Provider logs and backups may retain information longer.

Files you selected through ChatGPT were already provided to that service. OpenAI’s handling of those files, conversations, and returned reports or links is separate from this application’s processing. Saving or sharing a cleaned copy creates a copy controlled by you or the destination you choose.

3. Account and connection information

The hosted service requires a separate account email address and a high-entropy hosted-service access key. It stores the email, account identifier, a hash of the key, grant type, access-expiry date, request limit, and disabled status in Neon. Access-key and OAuth-token hashes are used for verification. For a new purchase or top-up, an access-key copy is encrypted using a secret, browser-bound claim for the receipt; the receipt claim expires after seven days and its ciphertext is cleared on later cleanup. Recovery clears old claims immediately. Full keys are not stored in plaintext in the application database.

Test grants and paid grants are distinct. Preview test access does not establish paid access. An Apple app purchase or ChatGPT plan does not automatically include this service. The website purchase flow uses Stripe. The application stores minimal order and payment references, amount, currency, payment/refund/dispute state, pack credits and counters, and purchase/expiry times. It does not receive card numbers or card security codes. A purchase record also retains the accepted Terms and Privacy versions, document hashes, acceptance timestamp, and method.

A browser-bound connection flow and its consent state last five minutes. The connection uses a necessary, secure, HttpOnly browser cookie for that flow. Authorization codes last two minutes. Access tokens last fifteen minutes. Refresh-token families last up to thirty days, with refresh rotation and replay detection. Revocation, an expired account grant, or a disabled account prevents further authorized service use. Expiry limits availability; physical row deletion follows the cleanup described above.

The account identifier and email are returned to ChatGPT when it requests the connected account profile. Request counters, processing leases, hashed token records, and expiry times are also used to enforce access, usage, concurrency, and download limits.

Account recovery sends a 15-minute single-use link to the existing account email through Resend, once the Company’s sending domain is verified. The token appears in the email link’s fragment and is not sent to the recovery-page server as a query parameter. Only its hash and a current account/credential snapshot are persisted for redemption. Confirming recovery replaces the key and invalidates existing connections, held jobs and temporary downloads. It does not extend a purchase or add credits, including for an unpaid or expired account record.

Necessary browser cookies protect account connection, website checkout receipts and recovery forms. The service uses salted hashes of the platform-reported requester address for recovery rate limits; it does not persist the raw address for that application counter. Request counters and single-use recovery proofs expire through the described cleanup. An encrypted receipt for a completed cleaning may retain a temporary download capability and generic report for up to ten minutes, so an identical request can recover a previously confirmed output.

4. Logs, analytics, and service providers

The hosted application uses no document-model calls, product analytics, advertising, or cross-app tracking. It does not write document contents, author values, or filenames to application logs. Unlike the iPhone, iPad, and Mac Catalyst native apps, this hosted service does not send product events to PostHog.

Vercel processes the service’s network requests and runtime operations; Neon stores account, connection, usage, and encrypted-output records. Stripe processes checkout and payment information. Resend processes the destination email and recovery message, including the secret recovery link, to deliver that message. Provider records and logs may remain longer than application-token expiry. Hosting providers may process IP addresses, request timing, request paths, errors, and other operational records. A request path can include the secret temporary download capability. We do not promise that provider logs contain no download URLs or that provider logs and backups are immediately or permanently erased.

We use information to operate authorized processing and downloads, enforce limits, maintain security and reliability, answer support requests, comply with law, and establish or defend legal claims. We do not use document contents for advertising, AI training, or product development.

We disclose information reasonably necessary to the service providers described here; to recipients when reasonably necessary to comply with valid process, protect rights or safety, investigate abuse, or establish or defend claims; and in a financing, reorganization, merger, acquisition, or sale subject to appropriate protections. Your use of ChatGPT or another file destination is also handled by that independent provider.

5. Support and retention

If you email us, we receive your email address, message, and anything attached. Do not email sensitive documents, access keys, access tokens, signed input URLs, or cleaned-copy download links. Remove private information from screenshots.

Account records remain available for administration and support until removed. Expired grants may remain as records but do not authorize use. Support correspondence is retained as reasonably needed to respond, maintain business records, resolve disputes, and meet legal obligations. No fixed support-record deletion period is promised.

Disconnecting ChatGPT revokes the connection when its revocation request reaches the service; it does not itself delete every account record, support email, provider log, backup, or copy already downloaded. You may request account deletion or access revocation at dean@myworkingmemory.ai. We may need to verify the request. Purchase and refund records may be retained for transaction administration, accounting, disputes and support. Recovering or deleting account access does not itself reverse a purchase. The hosted Terms describe the refund process.

6. Sale, advertising, and tracking

We do not sell personal information, use targeted advertising, share information for cross-context behavioral advertising, or track you across other companies’ apps or websites. The necessary connection cookie is not an advertising cookie. These pages themselves contain no tracking code.

7. Security and your choices

The service uses HTTPS, hashed access credentials, browser-bound account connection, scoped and expiring tokens, entitlement checks, encrypted temporary outputs, separate copies, and processing limits. No storage or transmission method is completely secure. Keep independent backups, protect access keys and temporary links, inspect first, and review every cleaned copy before sharing it.

You choose which file to select and whether to request a cleaned copy. Removing tracked changes, hidden text, or custom XML requires explicit content-change confirmation. You may disconnect the account in ChatGPT or request revocation from support. You control copies you download or share and must remove those separately.

Depending on your location and applicable law, you may request access, correction, deletion, or a copy of covered personal information, object to or restrict certain processing, or appeal a denied request. We do not discriminate for exercising applicable rights. Email dean@myworkingmemory.ai. We may need to verify the request.

We do not sell personal information or use targeted advertising, so there is no sale or targeted-advertising opt-out to exercise. Authorized agents may submit requests where applicable law permits. We will explain any lawful denial and available appeal process.

8. Children

The Services are general-audience tools and are not directed to children under 13. We do not knowingly collect personal information from a child under 13. If you believe a child supplied personal information through the Services or support, contact us so we can evaluate and delete it as required.

9. International users

Hosting, database, or support systems may process information in the United States or other locations where providers operate. Those locations may have different data-protection laws. Where required, we use lawful transfer and contractual safeguards appropriate to the information and provider relationship.

10. Changes and contact

We may update this policy when the product, providers, law, or practices change. The effective date above identifies this version.

Privacy questions, requests, or complaints: Myworkingmemory LLC, dean@myworkingmemory.ai.