Skip to content

Privacy Policy

Last reviewed: November 5, 2026

This page explains how personal data is processed when you visit the Published website or use account-related features. It documents all third-party services that process personal data on our behalf or in connection with our services. The application itself launches shortly and is already described here in full, so that you can see before signing up what will happen to your data; the sections on accounts, media, AI analysis, publishing, and payments apply from the moment it goes live.

1. Controller

Cosmic Code GmbH

Ehrenbergstraße 44, 22767 Hamburg, Germany

[email protected]

2. Purposes and legal bases

We process personal data only to the extent necessary to operate the website, provide secure account access, answer inquiries, and protect the product against misuse.

Depending on the specific processing activity, the legal bases are Article 6(1)(b) GDPR where processing is necessary to provide account and contract-related functionality, Article 6(1)(f) GDPR for security, abuse prevention, and technical operation, and Article 6(1)(c) GDPR where statutory retention duties apply.

3. Website access data

  • When you visit the site, technical request data may be processed, including IP address, browser and device information, timestamps, referrer, and requested resources.
  • We use this data to deliver the site securely, diagnose technical issues, and maintain system integrity.
  • Retention depends on operational necessity and security requirements.

4. Cookies and local storage

  • This website does not use tracking, analytics, or marketing cookies, and no third-party tracking scripts are loaded. A cookie consent banner is therefore not required.
  • If you switch between light and dark mode, your preference is stored in your browser’s localStorage. This storage is triggered by your own action and is strictly necessary to provide the function you requested; it is therefore permitted without consent (Section 25(2) no. 2 of the German TDDDG).
  • Our hosting provider Cloudflare may set strictly necessary cookies (for example “__cf_bm”) for bot detection and to protect the site against automated abuse. These cookies are required for the secure operation of the site (Section 25(2) no. 2 TDDDG, Article 6(1)(f) GDPR) and expire after a short time.

5. Desktop plugins (Lightroom, Eagle)

  • Our plugins for Adobe Lightroom Classic and Eagle run on your own computer and upload media into your account. They are our own software; no third party is involved in the connection itself.
  • Pairing runs through a sign-in in your browser plus a short-lived PIN. The PIN record holds the pairing state, your user and account reference, the number of attempts, and an expiry timestamp, and it is deleted once it has been redeemed or has expired.
  • A successful pairing issues an access token for the plugin. We store only a hash of that token plus a short non-secret prefix that lets you recognise it in the list, together with the time it was last used. You can revoke a token at any time in the settings, which ends the plugin’s access immediately.
  • Uploading transfers the original image and video files together with their embedded metadata — EXIF, XMP, and IPTC, which can include capture time, camera and lens, GPS coordinates, and the names of photographers or rights holders. This data is stored with the file and is used for the features that build on it, such as location tagging.
  • Legal bases: Article 6(1)(b) GDPR for the upload and the connection itself, and Article 6(1)(f) GDPR for the abuse protection built into pairing (attempt counting, lockout, expiry).

6. Profile pictures

  • People and brands in your account can show a picture: one from our gallery, an own upload, or — for people — the profile picture of the Google account you sign in with.
  • An own picture is cropped to a square in your browser and saved as a new image file without embedded metadata (such as EXIF location or camera data). It is stored with Cloudflare R2 under a public address that cannot be guessed.
  • The picture is visible to the people of your brands, to invited reviewers, and in e-mails to these people.
  • When you choose another picture or remove it, the deletion of the previous file is triggered.
  • Legal basis: Article 6(1)(b) GDPR — showing who works on which post is part of the team features you use.

7. AI agents (MCP connections)

  • You can connect an AI agent of your choice — for example Claude Code or Cursor — to your account through the Model Context Protocol (MCP). The agent reads data of the brands you release for it, creates drafts, uploads and projects, and proposes schedules, sets, captions and replies to comments. It cannot publish or schedule anything itself: every change that goes out is accepted by a person in the app.
  • The connection is created in the settings by you and belongs to you. The access token is shown only once; we store a hash of it and a short non-secret prefix, the name you gave the connection, the rights per brand you granted, the name of the agent software last reported, and the time of last use. The rights never go beyond your own role in a brand. You can change them or revoke the connection at any time; access then ends immediately.
  • Each call of the agent is logged: user, brand, tool, a short description, references to the affected posts or media, the result and, if any, an error code, plus the model and effort the agent software reports about itself (we do not verify these). Proposals are stored with their content (times, caption before and after, set media, reply text), the agent’s reasoning, and who decided on them.
  • Recipients: what the agent reads is sent by the agent software you use to its AI provider. This transfer is made by your agent software under your agreement with that provider, not by us, and it can include content of third parties in your brands, such as comments. Which provider that is depends on your agent, not on Published.
  • Storage periods: the log of calls 30 days; open proposals expire after 24 hours, decided and expired ones are deleted 30 days later; requests to open a page in the app after 24 hours. Deleting a brand removes its log and proposals.
  • Legal bases: Article 6(1)(b) GDPR for the connection and the functions you use through it, and Article 6(1)(f) GDPR for the log, which lets you trace what an agent did and protects against misuse.

8. Approval by reviewers

  • The holder of an account, for example an agency, can invite reviewers to a brand by email address. Reviewers approve posts before they are published or request changes. They do not need an account of their own: they sign in with a one-time code of six digits or a sign-in link, both sent by email.
  • We store the email address of the reviewer, the decisions (approved, changes requested) and comments with their times, and in the history of the post the name or email address of the person who acted. After sign-in the browser keeps a session token in its localStorage; on our side we store only a hash of it, likewise of the code and the sign-in link.
  • For this data we act on behalf of the account holder who invites the reviewers (Article 28 GDPR); the account holder is the controller and determines the legal basis. We limit how often codes can be requested and checked to protect the sign-in against misuse (Article 6(1)(f) GDPR).
  • Recipients: Brevo sends the emails (the code email and the summary of posts awaiting approval), Convex stores the data; both are listed under “Recipients” below.
  • Storage periods: code and sign-in link 10 minutes, usable once; the session 90 days or until sign-out. Removing a reviewer ends access immediately; the reviewer entry with the email address remains for as long as the brand exists, so that earlier decisions stay attributable. Decisions, comments and the history are kept for as long as the post exists.
  • Reviewers without an account can also request access to and erasure of their data, by email to [email protected]. After erasure the history of the post shows “Deleted reviewer” in place of the name.
  • Team invitations work the same way: an account holder or admin can invite colleagues by email address. We store the address, role, brands and times of the invitation; the link in the email is stored only as a hash and is valid for 7 days. Brevo sends the invitation email.

9. Email signup and opt-in proof

  • When you join the waitlist or newsletter, we store in our own database — as the controller, separately from our email provider — your email address, opt-in status, the request and confirmation timestamps, and the version of the consent text shown to you.
  • For abuse prevention (Article 6(1)(f) GDPR) we additionally record the IP address and user agent captured at opt-in. These two fields are automatically deleted 90 days after confirmation.
  • The opt-in status, timestamps, and consent copy version are retained as proof of consent (Article 7 GDPR) for as long as you remain subscribed.

10. Feedback on the help pages

  • On the help pages and in the help sidebar of the app you can tell us whether a guide was helpful, what was missing, and which guide you would like to have.
  • From the website we store only your answer (yes/no), the optional text, the guide, the language, and the time — without an account, without an IP address, and without cookies. In the app we additionally store your user reference so that the same person does not submit endlessly.
  • Please do not enter personal data in the free-text field. We use the feedback to improve the guides (Article 6(1)(f) GDPR) and delete it automatically after 12 months.

11. Processors and third-party services

The following third-party services process personal data on our behalf or in connection with our services.

11.1 Kinde — Authentication and user management

We use Kinde Pty Ltd (Level 4, 60 Martin Place, Sydney NSW 2000, Australia) for registration, sign-in, session management, user administration, and related security features. Kinde operates regional infrastructure; for our deployment customer data is processed in the European Union.

Under Kinde’s Data Processing Addendum, we act as controller for our end users’ personal data and Kinde acts as processor. Kinde may process limited account information as an independent controller where required to operate the service and the contractual relationship with us.

  • Categories of data: email address, name, profile image URL, authentication identifiers, session data, technical metadata such as IP address, browser and device information, and where applicable role and permission assignments.
  • Purposes: account creation, sign-in, verification, session management, abuse prevention, and account administration.
  • Legal bases: Article 6(1)(b) GDPR for account and contract functionality, plus Article 6(1)(f) GDPR where security and misuse prevention measures are required.

11.1.1 Transfer safeguards

Kinde is headquartered in Australia and runs regional data residency. EU customer data is hosted in EU regions. Where Kinde or its subprocessors process data outside the EEA, the Kinde DPA incorporates Standard Contractual Clauses as the transfer mechanism.

  • Primary measure: EU data residency for end-user authentication data.
  • Fallback mechanism: Standard Contractual Clauses as incorporated by the Kinde DPA.
  • Current subprocessors and their processing locations are published by Kinde on its trust and subprocessor pages.

11.1.2 Sign in with Google

Instead of an email code, you can sign in with your Google account. This is optional; signing in by email code works without Google. If you choose Google, you are redirected to Google’s sign-in page. There, Google Ireland Limited, Gordon House, Barrow Street, Dublin 4, Ireland, processes your sign-in as an independent controller under its own privacy policy.

After you confirm, Google passes the following to Kinde and from there to us: your name, your email address, your profile image and a Google account identifier. We do not request any other data from your Google account. We do not get access to your contacts, files or other Google services, and we do not post anything on your behalf.

  • Legal basis: Article 6(1)(b) GDPR, as the sign-in method you choose for using your account.
  • Google may also process data in the USA. Google LLC is certified under the EU-U.S. Data Privacy Framework (Article 45 GDPR).
  • You can remove the connection at any time in your Google account under “Third-party apps & services”. Your account with us remains, and you can sign in by email code.

11.2 Convex — Backend infrastructure and database

We use Convex, Inc., San Francisco, CA, United States, as our backend infrastructure provider. Convex hosts our application database, executes serverless functions, and provides file storage. In this context, Convex acts as a processor under Article 28 GDPR.

  • Categories of data: user profile data linked to authentication (user IDs, display names, email addresses), application content created or uploaded by users (documents, media references, project metadata, image metadata including EXIF/XMP/IPTC), organization and membership data, and technical metadata (timestamps, function execution logs).
  • Purposes: persistent storage of application data, execution of server-side business logic, real-time data synchronization between clients, and file storage.
  • Legal bases: Article 6(1)(b) GDPR for contract performance and Article 6(1)(f) GDPR for system integrity and security.

11.2.1 Processing location and transfer safeguards

Our Convex deployment runs in Convex’s European region. Convex hosts its platform on Amazon Web Services; the region we use is AWS eu-west-1 (Ireland). Application data is therefore stored and processed within the European Union.

Convex, Inc. is headquartered in the United States, so access from the United States for operating and supporting the platform cannot be excluded. Such access is governed by the Convex Data Processing Addendum, which forms part of our agreement with Convex and incorporates the EU Standard Contractual Clauses (Module Two, controller to processor) as the transfer mechanism.

11.3 Axiom — Log storage and error analysis

Our backend at Convex passes the logs of its server functions to Axiom, Inc., 1390 Market Street, Suite 200, San Francisco, CA 94102, United States. We use Axiom to store these logs, search them, and find the cause of errors. Axiom acts as a processor under Article 28 GDPR.

  • Categories of data: execution records of our server functions (function name, time, duration, status, error messages, resource usage) and the text our functions write to the log. This text can contain internal identifiers such as user, brand, and post IDs, as well as identifiers and names of connected social media accounts, for example the ID of a Facebook Page, a Threads username, or a platform account ID.
  • Purposes: monitoring the operation of the application, analysing errors, and investigating incidents and misuse.
  • Legal basis: Article 6(1)(f) GDPR. Our legitimate interest is a stable and secure service whose errors we can trace and fix.

11.3.1 Processing location, transfer safeguards and retention

Our Axiom datasets are located in Axiom’s edge deployment EU Central 1, which runs on Amazon Web Services in the region eu-central-1 (Frankfurt, Germany). Axiom stores the logs there, and queries and their results are processed there as well.

Axiom, Inc. is headquartered in the United States, so access from the United States for operating and supporting the platform cannot be excluded. When we analyse logs through Axiom’s hosted MCP server, Axiom routes the query results through its US infrastructure. Such transfers are governed by the Axiom Data Processing Addendum, which forms part of our agreement with Axiom and incorporates the EU Standard Contractual Clauses (Module Two, controller to processor).

Axiom deletes the logs automatically after 30 days, the retention period of our Axiom plan.

11.4 Cloudflare — Hosting, CDN, and object storage

We use Cloudflare, Inc., 101 Townsend Street, San Francisco, CA 94107, United States, for website hosting (Cloudflare Pages), serverless computing (Cloudflare Workers), content delivery (CDN), and object storage (Cloudflare R2). Cloudflare acts as a processor for these services.

  • Categories of data: IP addresses and technical request metadata (browser, device, timestamps, referrer) processed during content delivery, as well as user-uploaded media files stored in R2 (which may contain embedded image metadata such as EXIF data including GPS coordinates).
  • Purposes: secure delivery and hosting of the website, edge computing for image processing, and persistent storage of user-uploaded media.
  • Legal bases: Article 6(1)(b) GDPR for providing the service and Article 6(1)(f) GDPR for security, performance optimization, and DDoS protection.

11.4.1 Transfer safeguards

Our R2 bucket is created with the EU jurisdictional restriction. Cloudflare guarantees for such buckets that the objects they hold are stored within the European Union — user-uploaded media therefore rests exclusively on EU infrastructure.

Delivery is a separate matter from storage: hosting, CDN, and image processing run on Cloudflare’s global network, and content is served from the edge location closest to the visitor. Temporary cached copies of publicly delivered media can therefore exist at edge locations outside the EU.

Cloudflare, Inc. is based in the United States. It is self-certified under the EU–U.S. Data Privacy Framework and offers a Data Processing Addendum that includes Standard Contractual Clauses.

11.5 Stripe — Payment processing

We use Stripe, Inc., 354 Oyster Point Boulevard, South San Francisco, CA 94080, United States, and its EU entity Stripe Payments Europe, Ltd., 1 Grand Canal Street Lower, Dublin 2, Ireland, for payment processing. Depending on the processing activity, Stripe acts as a processor on our behalf or as an independent controller for its own fraud prevention, regulatory compliance, and financial reporting obligations.

Payment card details are collected directly by Stripe through Stripe Elements. Card numbers are never transmitted through or stored on our servers. Stripe is a certified PCI DSS Level 1 service provider.

  • Categories of data: name, email address, billing address, payment method details (handled directly by Stripe), transaction amounts, currency, transaction identifiers, and device data and IP address for fraud detection.
  • Purposes: processing payments for subscriptions or purchases, invoice generation, fraud detection and prevention, and regulatory compliance (anti-money laundering, sanctions screening).
  • Legal bases: Article 6(1)(b) GDPR for payment processing necessary to perform the contract, Article 6(1)(c) GDPR for Stripe’s statutory obligations as a payment service provider, and Article 6(1)(f) GDPR for fraud prevention.

11.5.1 Transfer safeguards

Stripe Payments Europe, Ltd. processes EU payment data within the EEA where possible. For transfers to Stripe, Inc. in the United States, Stripe relies on the EU–U.S. Data Privacy Framework and Standard Contractual Clauses.

11.6 Google Vertex AI — AI-assisted image analysis, text generation, and comment sorting

We use Vertex AI, a Google Cloud service, for automated image and text analysis inside our application. For customers with a billing address in Europe, the Middle East, or Africa, the contracting entity is Google Ireland Limited, Gordon House, Barrow Street, Dublin 4, Ireland. Google acts as a processor under the Cloud Data Processing Addendum.

Note that this is the enterprise Google Cloud offering, not the Gemini consumer app or the Gemini Developer API — a different set of terms applies, and Google does not use the data we submit to train or improve its models.

  • Categories of data: uploaded image files and video frames, their embedded metadata, captions and other text submitted for analysis or generation, the text of comments under a brand’s posts if the brand has switched on AI in the comment inbox, the analysis results themselves, and the API request data required to authenticate the call.
  • Purposes: structured analysis of uploaded media (among others mood, style, shot type, lighting, colour, subject, and context), generation of caption and hashtag suggestions per target platform, sorting comments in the comment inbox by topic and drafting reply suggestions, and the calculation of vector embeddings that make semantic search across a media library possible.
  • Legal bases: Article 6(1)(b) GDPR, because AI-assisted analysis and generation are core functions of the product, and Article 6(1)(f) GDPR for assessing content quality and relevance.
  • No automated decision-making with legal or similarly significant effect within the meaning of Article 22 GDPR takes place. Every generated result is a draft that users review, edit, or discard.

11.6.1 Comments in the comment inbox

AI in the comment inbox is off by default. An owner or admin of a brand can switch it on for that brand; we record who did so and when. Simple cases, such as praise made only of emoji or comments the platform has marked as spam, are sorted by fixed rules without any transfer. Only the remaining comments go to Vertex AI.

The comment text goes out as it was written and may contain names or handles. We add no metadata: no handle, profile picture, or identifier of the person who commented. Alongside it we send the caption of the post and the text of the brand profile, so that suggestions fit the brand. The model returns a topic for the comment (request, question, criticism, praise, or none) and, for requests and questions, three reply suggestions.

Comment sorting runs only on the EU endpoint. Suggestions are drafts: nothing is posted automatically, and a user decides whether to send, edit, or discard them. A topic belongs to the comment, not to the person who wrote it. Suggestions are deleted when the comment text changes, when the AI is switched off, and together with the comment. When a user asks to improve a reply or for new suggestions, the result is shown but not stored.

11.6.2 Processing location

We call Vertex AI in the Google Cloud regions europe-west4 (Netherlands) and the EU multi-region. Analysis and generation therefore run on European infrastructure.

Google Ireland Limited is established in the European Union. Where Google or its affiliates access data from outside the EEA — for support and operations, for example — the Cloud Data Processing Addendum provides the transfer mechanism, including Standard Contractual Clauses.

11.6.3 Abuse monitoring

Google documents that, where automated safety classifiers flag suspicious activity, prompts may be logged in order to examine whether the acceptable use policy has been violated. According to Google, such logged data is not used to train models, is stored for up to 90 days, and remains in the region or multi-region selected for the project — for us, therefore, within the EU.

11.7 Social media integrations — Publishing to social media platforms

Our application publishes and manages content through accounts on Instagram, Facebook, Threads, X, Bluesky, Pinterest, TikTok, YouTube, LinkedIn, Mastodon, and Pixelfed; which of these platforms can currently be connected is shown in the app. When a user connects a platform, our application accesses that platform’s API on the user’s behalf using OAuth tokens.

The platforms themselves act as independent controllers for the data processed on their own services — including the data of people who see or interact with published content. We act as a controller for the OAuth tokens and connection data stored in our systems. OAuth tokens are stored encrypted.

  • Categories of data: OAuth access and refresh tokens (stored encrypted), platform user identifiers, page, board, or profile identifiers, profile names and avatars, platform-specific content identifiers, and the content and performance metadata returned by the platform APIs; with the comment inbox switched on, also the comments under the user’s posts (see “Comment inbox”).
  • Purposes: authenticating with the platform on the user’s behalf, publishing content to connected accounts, retrieving post performance data, retrieving comments on published posts and sending the replies and likes a user chooses, and managing platform connections.
  • Legal bases: Article 6(1)(b) GDPR for providing the publishing functionality as part of the service.

11.7.1 Publishing to a platform is a transfer to that platform

Whatever a user publishes through us — images, videos, captions, hashtags, and any location a user tags — is transmitted to the selected platform and is subject to that platform’s own privacy policy and terms from that point on. Several of these platforms are based in the United States and operate globally. Users decide themselves which content goes to which platform.

Where an image shows identifiable people, publishing it makes their personal data available on the platform in question. Responsibility for having the necessary legal basis for that publication lies with the user who publishes.

11.7.2 Mastodon and Pixelfed (Fediverse)

Mastodon and Pixelfed are not run by one company. The user chooses the server (instance) their account lives on; the operator of that server is the independent controller for what is processed there, under that server’s own privacy policy. We connect only to the server the user enters.

For each server we register our application once and store its client credentials (the secret encrypted), plus the server address, the account identifier, and the limits the server publishes (for example the maximum text length). Published content is distributed by the user’s server to other servers in the Fediverse (ActivityPub); copies on those servers are outside our and the original server’s control, and deleting a post does not guarantee that every copy is removed.

When a user disconnects a Mastodon account, we also ask the server to revoke our access token. Pixelfed offers no way for us to do that, so the app asks the user to revoke our application in their server’s settings.

11.7.3 YouTube (YouTube API Services)

Our application uses YouTube API Services to connect YouTube channels, upload videos, show their statistics, and, with the comment inbox switched on, read and answer the comments under those videos. By connecting a YouTube channel, users agree to the YouTube Terms of Service. Google’s Privacy Policy applies to the processing by YouTube.

Through the YouTube API Services we access the channel identifier, the channel name and avatar, the identifiers and settings of the videos uploaded through us, their statistics and YouTube Analytics data such as watch time, and the comments under those videos with the commenter’s display name, handle, and profile picture URL. We use this data only to provide these functions in the user’s account. Apart from YouTube itself, it reaches only the processors named in this policy: Convex for storage and, if a brand switches on AI in the comment inbox, Google Vertex AI for the comment text.

Stored YouTube API data is refreshed or deleted: data we have not confirmed through the API within the last 30 days is deleted. This includes comments, which we refresh through the API and delete at the latest 29 days after the last successful refresh. After a user disconnects the channel or revokes our access, we delete the stored YouTube API data within 7 days. Users can revoke our access at any time in the Google security settings. We do not store or read information on users’ devices through the YouTube API Services.

11.7.4 LinkedIn

Our application uses LinkedIn’s Community Management API. A brand connects either the LinkedIn profile of the connecting member or one LinkedIn Page that member manages. We publish the posts the user approved there and show their statistics. Through the API we access the identifier, name, and picture of the connected member, the identifiers of the Pages they manage with the Page name and logo, the identifiers of the posts published through us, and aggregated statistics: per post (for example impressions and reactions) and the follower count of the Page or profile.

With one exception, we show no data of other LinkedIn members: for posts on a connected Page, the brand’s admins can open the comments under the post. We load them live from LinkedIn, show the comment text and time, and do not store them. Statistics are aggregated counts for the user’s own posts, visible only to the members of the brand that owns the connection. We offer no export of LinkedIn data.

Storage periods follow LinkedIn’s rules. Page name, logo, and current follower count are deleted if we have not confirmed them through the API within 8 weeks. Statistics and follower history are deleted one year after they were measured at the latest. The profile of the connected member and the identifiers are kept like those of other platforms (see “Token lifecycle and disconnection”). Users can revoke our access at any time in their LinkedIn settings under “Data privacy” → “Permitted services”.

11.7.5 Comment inbox

If a brand switches on the comment inbox, our application retrieves the comments under posts that were published through us or imported, from Instagram, Facebook, Threads, Bluesky, Mastodon, Pixelfed, and YouTube, and shows them in one list. Users read them, reply right away or at a chosen time, like them where the platform allows it, and mark them as done. A reply or like is sent only when a user triggers it.

For these comments we act as a processor for our customer: the brand that receives a comment decides what happens with it and is the controller towards the person who wrote it. If you commented on a brand’s post and have a question about your data, please contact that brand. Requests that reach us are forwarded to it.

  • How comments reach us: Meta notifies us of new Instagram and Facebook comments via webhooks, and we fetch replies through the Graph API. For Threads and YouTube we query the comments under each post at intervals. For Bluesky, Mastodon, and Pixelfed we read the notifications and the posts of the connected account.
  • Categories of data: comment text, display name, handle, the URL of the profile picture (not the picture itself), the time of the comment, the platform’s comment identifier and its place in the conversation; plus status, read time, topic, reply suggestions, reply drafts, and sent replies. We do not store the platform user IDs of people who comment.
  • Scope: we store only comments addressed to the connected account (directly under the post, or replies to its own replies) together with the path to them. Conversations between other people are loaded when needed and not stored.
  • Storage period: a comment is deleted 90 days after it was marked as done, an open or scheduled comment 90 days after it was written, and every comment no later than 120 days after it was written. YouTube comments can be deleted earlier (see the YouTube section). A comment deleted on the platform is removed once we detect the deletion; we check every stored comment at least every 7 days. The text of a sent reply is deleted right away. Switching the inbox off, disconnecting the account, or deleting the brand deletes the comment data concerned.
  • Legal basis: Article 6(1)(b) GDPR towards our customer. The brand determines the legal basis for processing the data of the people who comment.
  • We do not build profiles of people who comment and do not link them across platforms or brands. A topic belongs to a comment, never to a person.

11.7.6 Token lifecycle and disconnection

OAuth tokens are stored only for as long as a platform connection is active. When a user disconnects a platform, the access token — and, for Bluesky, the session — is deleted from our systems and the connection is marked as disconnected. Its platform identifier and profile name and the content already synchronised from it remain stored until the brand is deleted (with the exceptions listed in the deletion guide) or we have carried out an erasure request. Comment inbox data and YouTube API data are deleted on disconnection, as described in those sections. Token refresh cycles follow each platform’s expiry policy.

11.8 Dropbox — Importing media files from the user’s own Dropbox

Users can import images and videos from their own Dropbox into the media library. For this we use the Dropbox Chooser, a selection window provided by Dropbox. The script for it is loaded from dropbox.com only when the user clicks “From Dropbox”; until then, no connection to Dropbox is made. From that moment the user’s browser connects directly to Dropbox, which receives the IP address, browser information, and the address of our application. The user signs in and selects files in Dropbox’s own window. For this, Dropbox acts as an independent controller, and the user’s own agreement with Dropbox and Dropbox’s privacy policy apply.

We do not receive or store a Dropbox password or access token, and we do not keep a lasting connection to the Dropbox account. We see only the files the user selects. For each selected file, Dropbox hands our application the file’s Dropbox ID, its name and size, a direct download link valid for four hours, and a link to a preview image.

Our backend then fetches each selected file once through its download link and stores it in our own object storage (Cloudflare R2, EU jurisdiction). The download runs on Cloudflare’s infrastructure, not from the user’s browser. From there the file goes through the same processing as any other upload, including image analysis with Google Vertex AI. While the import is running, the user’s browser loads the preview images directly from Dropbox; for videos, a preview image may be stored as the video’s cover image.

  • Categories of data: the selected image and video files with their embedded metadata (as with any upload), file name and size, the Dropbox file ID, the time of selection, and an error code if an import fails. The download link is not stored with the asset; it is held only in our internal job data (the import queue and its run history), which deletes it no later than about seven days after the import. Dropbox invalidates the link after four hours.
  • Purposes: importing the files the user selected into the media library and showing where an asset came from (“Imported from Dropbox” in the asset’s info panel). The Dropbox file ID also prevents the same file from being imported twice.
  • Legal bases: Article 6(1)(b) GDPR, since the import is a function the user requests.
  • Storage period: the imported file, the Dropbox file ID, and the time of selection are kept with the asset and deleted together with it. If an import fails, we keep no asset; the file name and error code stay until the user dismisses the import in the activity panel, or until the brand is deleted. Exception: if a post already uses the file, the failed asset stays in the library with its error code until the user deletes it.
  • The import is a one-off copy. Later changes to the file in Dropbox are not carried over, and deleting the asset in Published does not delete the file in Dropbox.

11.9 Google Drive — Importing media files from the user’s own Google Drive

Users can import images and videos from their own Google Drive into the media library. When the user clicks “Import from Google Drive”, a short notice explains that the selected files are copied into the media library and analyzed there automatically, including by AI. Only then does the user’s browser load Google’s sign-in and file-picker scripts (accounts.google.com, apis.google.com, docs.google.com); before that click no connection to Google is made. From that moment Google receives the IP address, browser information, and the address of our application. Sign-in, consent, and file selection happen in Google’s own windows. For this, Google acts as an independent controller, and the user’s own agreement with Google and Google’s privacy policy apply.

We ask Google only for the permission “See, edit, create and delete only the specific Google Drive files you use with this app” (drive.file). It covers the files the user picks for us, not the rest of the Drive. We do not receive a Google password or a refresh token, and we do not keep a lasting connection to the Google account. For each selected file, the picker hands our application the Drive file ID, its name, size and file type, together with an access token that Google lets expire after at most one hour.

Our backend then fetches each selected file once from the Google Drive API with this token and stores it in our own object storage (Cloudflare R2, EU jurisdiction). The download runs on Cloudflare’s infrastructure, not from the user’s browser. From there the file goes through the same processing as any other upload, including image analysis with Google Vertex AI. We do not load preview images from Google Drive.

Our use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. We use Drive data only for the import the user started and for the functions of the media library the user sees. We do not sell it, do not use it for advertising, and do not use it to train general AI models.

  • Categories of data: the selected image and video files with their embedded metadata (as with any upload), file name, size and type, the Drive file ID, the time of selection, and an error code if an import fails. The access token is not stored with the asset; it is held only in our internal job data (the import queue and its run history), which deletes it no later than about seven days after the import. Google invalidates it after at most one hour.
  • Purposes: importing the files the user selected into the media library and showing where an asset came from (“Imported from Google Drive” in the asset’s info panel). The Drive file ID also prevents the same file from being imported twice.
  • Legal bases: Article 6(1)(b) GDPR, since the import is a function the user requests.
  • Storage period: the imported file, the Drive file ID, and the time of selection are kept with the asset and deleted together with it. If an import fails, we keep no asset; the file name and error code stay until the user dismisses the import in the activity panel, or until the brand is deleted. Exception: if a post already uses the file, the failed asset stays in the library with its error code until the user deletes it.
  • The import is a one-off copy. Later changes to the file in Google Drive are not carried over, and deleting the asset in Published does not delete the file in Google Drive. The user can remove our access at any time in their Google account (Security → Third-party apps with account access).

11.10 Canva — Importing, editing and creating designs with the user’s own Canva account

Users can connect their own Canva account to Published to import designs as images into the media library, to keep editing imported designs in Canva, and to start new Canva designs from their own photos. The connection belongs to the individual Published user, not to the brand. It is made only when the user chooses “Connect Canva” in the settings and agrees on Canva’s own page (OAuth 2.0 with PKCE). The user grants these permissions: read designs and their metadata, read and create design content, upload images to their Canva uploads and read them, and read their profile. The data in the Canva account is governed by the user’s own agreement with Canva and by Canva’s privacy policy; Canva acts there as controller, or as the customer’s processor where the customer has Canva’s Data Processing Addendum.

Roles: for the connection (tokens, Canva user and team ID, display name) and the technical records described below, we are the controller. Imported designs and the photos a user sends to Canva are our customers’ content; for these we act as processor on the customer’s behalf, as for every other upload. Whether the customer may send a photo showing other people to Canva, and on what legal basis, is the customer’s decision as controller.

When the connection is made, Canva gives us an access token and a refresh token, the Canva user ID, the Canva team ID and the display name of the Canva profile. We store the tokens only in encrypted form (AES-GCM) in our database (Convex); they do not appear in logs. Our backend uses them to call the Canva Connect API, and only when the user uses one of the Canva functions or when an export the user started is running.

Import: the “From Canva” dialog shows the user’s designs with title, page count, time of last change and a preview image; a search term the user enters is sent to Canva. The user’s browser loads the preview images directly from Canva, which receives the IP address and browser information. For the selected pages our backend asks Canva for a PNG export. Canva returns time-limited download links; our backend fetches each page once on Cloudflare’s infrastructure and stores it in our own object storage (Cloudflare R2, EU jurisdiction). From there the image goes through the same processing as any other upload, including image analysis with Google Vertex AI. Each page becomes its own asset, named after the design title and the page number.

Edit in Canva: this opens the design in Canva’s editor, on Canva’s website, in the same browser tab. For the way back we keep a record of the trip (design ID, Canva user and team ID, the original asset, the app page to return to), which expires after 24 hours. On return Canva hands over a signed token with the Canva user ID, team ID and design ID; we check its signature with Canva’s public keys. We import the changed pages as a new version in the same way as an import; the original asset stays as it is.

Design in Canva: only when the user starts this for a photo do we send that photo to the user’s Canva account. We do not send the original but our working copy (WebP, at most 2,000 pixels on the longer side), together with the asset name as the upload name (shortened to 50 characters) and as the design title. Our Cloudflare Worker reads the file from R2 and streams it directly to Canva. The photo then remains in the user’s Canva account as an upload and as the basis of a new design. When the user returns, the new design is imported as described above.

  • Categories of data: Canva user ID, Canva team ID, display name of the Canva profile, granted permissions, access and refresh token (encrypted), time of connection; for each imported page the design ID, page number, design title (as asset name), time of selection and of the design version, and an error code if an import fails; the exported images; for “Design in Canva” the working copy of the selected photo and its name. Temporary: the export download links (held only in our internal job queue until the image has been fetched), the state of the sign-in handshake (single-use, 5 and 10 minutes), and per-Canva-user counters for Canva’s rate limits.
  • Purposes: connecting the user’s Canva account, importing the designs the user selects into the media library, opening imported designs in Canva and taking back the changed version, starting new designs from the user’s photos, and showing where an asset came from. The design ID and page also prevent the same page from being imported twice.
  • Legal basis: for the data we are controller of, Article 6(1)(b) GDPR, since each of these is a function the user requests. For customer content, the customer as controller determines the legal basis.
  • Storage period: tokens, Canva user and team ID and the display name are kept until the user disconnects Canva in the settings or their account is deactivated. Disconnecting deletes them immediately and withdraws our access at Canva. Records of open edit trips expire after 24 hours and are removed by an hourly clean-up. The origin data of an imported asset (design ID, page, Canva user and team ID, times) is kept with the asset and deleted together with it. If the import of a page fails, we keep no asset; its title and error code stay until the user dismisses the import in the activity panel, or until the brand is deleted. Exception: if a post already uses the page, the failed asset stays in the library with its error code until the user deletes it. The record of an export job (Canva user and team ID, design ID, pages, status, error code) currently has no fixed deletion period; it is deleted when the account is deleted.
  • Disconnecting Canva or deleting an asset in Published does not delete anything in Canva. Designs, and photos uploaded through “Design in Canva”, stay in the user’s Canva account until the user deletes them there. Imported assets stay in Published after disconnecting.

11.11 OpenStreetMap (Nominatim) — Location search for post tagging

When a user searches for a place to tag on a post, or when we turn the GPS coordinates found in an image’s metadata into a place name, we query Nominatim, the geocoding service of the OpenStreetMap Foundation, St John’s Innovation Centre, Cowley Road, Cambridge CB4 0WS, United Kingdom. The Foundation acts as an independent controller for the operation of that service.

The query is sent from our backend, not from the user’s browser. The user’s IP address and browser information are therefore not disclosed to the Foundation; what reaches it is the search term or the coordinates.

  • Categories of data: the search term entered by the user, or latitude and longitude — which may originate from the GPS data embedded in an uploaded photo.
  • Purposes: finding places for location tagging and resolving coordinates into a readable place name.
  • Legal bases: Article 6(1)(b) GDPR, since location tagging is part of the publishing functionality the user requested.

11.11.1 Use by the Foundation and transfer safeguards

According to its privacy policy, the Foundation may analyse Nominatim queries in order to improve the OpenStreetMap dataset, for example to identify missing addresses and postcodes, and may publish data derived from this. The Foundation states that it does not release personal information, or information linked to an individual, to third parties.

The United Kingdom is a third country for which the European Commission has issued an adequacy decision (renewed on 19 December 2025 and valid until 27 December 2031). The transfer is therefore based on Article 45 GDPR and requires no additional safeguards.

11.12 Brevo — Email delivery and contact list management

We use Brevo (operated by Brevo SAS, Paris, France; DACH entity Sendinblue GmbH, Cologne, Germany) to deliver our emails and to manage subscriber contact lists. Brevo acts as a processor under Article 28 GDPR.

We send two kinds of email over separate streams, and they rest on different legal bases.

  • Categories of data: email address, list membership, the content of the email sent, and delivery events (delivery, opens, clicks, unsubscribes, bounces).

11.12.1 Marketing emails

  • Content: the double opt-in confirmation email, waitlist announcements, and the newsletter.
  • Legal basis: Article 6(1)(a) GDPR — consent, obtained by double opt-in and revocable at any time via the unsubscribe link in every email.

11.12.2 Transactional emails

  • Content: notifications about your own use of the product — reports on whether scheduled posts published successfully, and warnings when the connection to a social media account has expired or failed.
  • These emails carry the data needed to make the notification useful: the affected brand and social account, post status and scheduling times, and the reason a publication failed.
  • Legal bases: Article 6(1)(b) GDPR, because these notifications are part of the service, and Article 6(1)(f) GDPR for informing users about faults in their connected accounts.
  • They are not marketing and therefore carry no unsubscribe link. Which notifications you receive is configured in your account settings.

11.12.3 Transfer safeguards

Brevo is an EU-based provider and processes data on servers within the European Union. No transfer to a third country is required for this service.

11.13 webhostone — Mailbox hosting for our contact addresses

The mailboxes behind our published.social contact addresses are hosted by webhostone GmbH, Rheinvogtstraße 17, 79713 Bad Säckingen, Germany. The provider receives and stores email sent to us and dispatches the replies we write. webhostone acts as a processor under Article 28 GDPR.

This is separate from Brevo, which we use to send emails to our subscribers and users. Anything you write to us directly lands here.

  • Categories of data: your email address and sender name, the subject and body of your message, any attachments, and the technical headers of the message (timestamps, the IP address of the sending mail server, delivery paths).
  • Purposes: receiving, storing, and answering enquiries sent to our contact addresses.
  • Legal bases: Article 6(1)(b) GDPR where the correspondence concerns a contract or steps taken prior to entering into one, Article 6(1)(f) GDPR for all other enquiries, and Article 6(1)(c) GDPR where commercial correspondence is subject to statutory retention duties.

11.13.1 Storage location and retention

The mail servers used are operated in Germany, so no transfer to a third country takes place for this service.

We keep correspondence for as long as it is needed to deal with your enquiry and any follow-up questions. Where an email qualifies as commercial or business correspondence, statutory retention periods under German commercial and tax law apply (six or ten years, Section 257 HGB, Section 147 AO).

12. International transfers

The processing that touches the most data runs inside the European Union: application data and file storage at Convex in Ireland, AI analysis at Google Vertex AI in EU regions, authentication at Kinde in EU regions, email delivery at Brevo on EU servers, and the operating logs at Axiom in Frankfurt.

Third-country transfers remain in three places. Cloudflare serves the site from the edge location closest to the visitor and therefore operates globally. Stripe processes payment data in the United States. Content that a user publishes goes to the social media platform they selected, several of which are US-based. In addition, access from the United States cannot be excluded at Convex and Axiom, and when we analyse logs through Axiom’s hosted MCP server, the query results pass through Axiom’s US infrastructure. A user who starts “Design in Canva” sends that photo to their own Canva account at Canva Pty Ltd in Australia (no adequacy decision); we carry this out as processor on the customer’s instruction, under the customer’s own agreement with Canva (see the Canva section above). Where personal data is transferred to a third country, the transfer relies on the mechanism described in the relevant service section above (EU–U.S. Data Privacy Framework, Standard Contractual Clauses, or other applicable safeguards).

13. Storage periods

Account data and the content you upload are stored for as long as your account exists.

We do not currently run an automated account-deletion routine. If you ask us to erase your data, we carry out the deletion by hand and inform you of the outcome within the period set by Article 12(3) GDPR — one month from receipt of the request, extendable by a further two months for complex requests, in which case we will tell you within the first month.

  • Deleted automatically on a schedule: the IP address and user agent recorded at newsletter opt-in, 90 days after confirmation; email delivery records after 30 days; notification events after 45 days; the pairing PIN for a desktop plugin after five minutes; OAuth state tokens after a short period.
  • Deleted on your action: the access token of a desktop plugin or an agent connection when you revoke it — access ends immediately, the stored hash and prefix are deleted 30 days later — and the access token (for Bluesky also the session) of a social media account when you disconnect it. The connection itself is marked as disconnected; its platform identifier and profile name and the content already synchronised from it remain stored and are removed together with the brand — with the exceptions listed in the deletion guide — or on request.
  • Retained despite an erasure request: data subject to statutory retention duties, in particular invoices and commercial correspondence under Section 257 HGB and Section 147 AO (six or ten years). Its processing is restricted to those retention purposes.
  • Log and security data is kept for short operational periods where this is necessary to detect abuse or investigate incidents.

14. Recipients

Personal data is disclosed only where this is necessary for service delivery, legal compliance, or secure technical operations. The following recipients process personal data in connection with our services:

A current list, stating for each provider its legal entity, where it processes data, and on what basis, is kept at published.social/subprocessors.

  • Kinde for Authentication and user management.
  • Convex for Backend infrastructure and database.
  • Axiom for Log storage and error analysis.
  • Cloudflare for Hosting, CDN, and object storage.
  • Stripe for Payment processing.
  • Google Vertex AI for AI-assisted image analysis, text generation, and comment sorting.
  • Social media integrations for Publishing to social media platforms.
  • Dropbox for Importing media files from the user’s own Dropbox.
  • Google Drive for Importing media files from the user’s own Google Drive.
  • Canva for Importing, editing and creating designs with the user’s own Canva account.
  • OpenStreetMap (Nominatim) for Location search for post tagging.
  • Brevo for Email delivery and contact list management.
  • webhostone for Mailbox hosting for our contact addresses.

15. Data subject rights

  • Right of access under Article 15 GDPR.
  • Right to rectification under Article 16 GDPR.
  • Right to erasure under Article 17 GDPR.
  • Right to restriction of processing under Article 18 GDPR.
  • Right to data portability under Article 20 GDPR.
  • Right to object under Article 21 GDPR.
  • Right to lodge a complaint with a supervisory authority.

16. Security

We implement technical and organizational measures appropriate to the risk in order to protect personal data against accidental or unlawful destruction, loss, alteration, unauthorized disclosure, and unauthorized access.

17. Updates to this policy

We may update this privacy policy where our services, processors, or legal obligations change. The version published on this page is the current version.

References