Privacy notice

This notice explains how Eclose processes personal data in its products — Eclose Omics, PubVia, Orasis and Eclose AI — as Articles 13 and 14 of the General Data Protection Regulation (EU) 2016/679 (“GDPR”) require.

Last updated: 2026-11-18

In short

This is the short version. The full notice below is the one that applies.

Who we are. Eclose runs Eclose Omics, PubVia, Orasis and Eclose AI. Write to main@eclose.bio with any question or request about your data.

Asking for access. The request form asks for your name, your email address, your operating system, the products you want and what you will use them for. Your institution is optional. We keep your request until every product you asked for is decided, then for up to 12 months.

Your account. We keep your user id and the movements of your balance (what you added and what you were charged) on our server in Amsterdam, Netherlands (EU). If you add a recovery email, we use it to send you the sign-in codes, new-token links and the other mails listed in section 4. You can remove it at any time.

Eclose Omics. You upload sequencing files. We analyse them on Scaleway machines in Paris, France (EU) and store the results there. We delete your sequencing files 4 days after upload; files uploaded before that rule are kept with the results. We keep your results for 365 days from the day the run starts for a run started now, or 30 days for a run started before (the date is shown on each run). Our record of each run (names, files, conditions) stays for at least 90 days, longer while a question about the job's machine is open, then shrinks to a billing summary. If you use an AI assistant, its provider receives what our server sends to it, under your own agreement with that provider.

PubVia. Your manuscripts stay on your computer, except what you send to your AI provider and the search services. When PubVia searches for papers, the search terms reach the search services, some through our server. We store none of the search terms; our server keeps only a coded form of your IP address for about 3 days, to limit requests. If you send an error report, we keep it for 30 days, in the EU.

Orasis. Your images and results stay on your computer, except what your AI client receives. We record each download of the app for 90 days, in the United States (section 7 explains the safeguards). If you send a problem report, we keep it for 30 days, in the EU.

Eclose AI. The chat in your account area, and Eclose Folio once it is released, send your requests through our server to an AI provider in the EU. If that provider fails, a fallback provider in the EU takes over. Both are our processors, and the request asks them not to store it and not to train on it. Folio sends file details, never whole files, except the text of the lab-book entry you work on and short excerpts of notes. In Folio's “your own AI” route, what your app reads goes straight from your computer to the provider you chose. It does not pass through us, except three tools that call Eclose AI. We delete a chat conversation 7 days after its last message. We keep a remembered fact for 12 months. Eclose Folio is not available yet.

Security and providers. We keep a security log for 12 months and remove the IP addresses from it after 90 days. We also keep the log of sign-ins to our protected areas, with the email and IP address, for 12 months, and traffic counts for 8 days (section 13). Cloudflare carries the traffic to our sites. Resend sends our emails and keeps its logs in the United States (section 7 explains the safeguards).

Your rights. You can ask for a copy of your data, a correction, its deletion, a restriction of its use, or a copy in a portable form, and you can object. You can also complain to a data protection authority. Section 9 explains how.

1. Who we are and how to contact us

Eclose is the controller of your personal data: we decide why and how it is processed (Art. 4(7) GDPR). In this notice, “we” means Eclose.

Write to main@eclose.bio for any request about this notice or your data.

2. What this notice covers

This notice covers four Eclose products: Eclose Omics, PubVia, Orasis and Eclose AI (the chat and Eclose Folio). Who we are, how to contact us, your rights, transfers outside the EEA and changes to this notice are the same for all four parts. The other sections have a separate part for each where the facts differ.

Eclose Omics

Eclose Omics is our service for analysing sequencing data: bulk RNA-seq, small RNA and single-cell RNA-seq. Only people whose access request we have approved can use it. In this notice, “the service” means Eclose Omics. This part covers:

It also covers requests you email to the contact address in section 1, such as a request for access. We process your email address and what you write. The legal basis is Art. 6(1)(b) GDPR (steps you ask us to take before an agreement) or, for other messages, Art. 6(1)(f). We keep these emails until the request is closed and for at most 12 months after that. If we create an account for you, the account's rules apply instead (sections 3, 4 and 8).

PubVia

PubVia is our writing app for macOS. It runs on your computer. Your manuscript, library, chat history and notes stay there. This part covers what the app sends from your computer, who receives it, and what our server for PubVia receives and keeps. That server is pubvia.eclose.bio, and it runs on Cloudflare. We do receive some of what you write. Online-search sentences pass through our server. The log in an error report can contain words from your work.

PubVia has no account. What the app sends us carries no account, no name or email address the app holds for you, and no identifier of your installation. Our server does see your computer's IP address, as every internet connection does. It keeps a keyed hash of that address for about 3 days, to limit searches. It never passes the address on. Any service the app contacts directly also sees your IP address.

Orasis

Orasis is our Windows app for analysing microscopy images on your own computer. It includes Fiji/ImageJ and the Bio-Formats library. Your images, files, lab book and results stay on your computer. The exception is what this part describes. By default, your AI client receives the names of your files and folders, and your lab-book and macro titles. Orasis has no account. This part covers what Orasis sends from your computer, and to whom. It also covers what our server for Orasis receives and keeps. That server is orasis.eclose.bio, and it runs on Cloudflare.

Requests for access to Orasis, and the “Report a problem” box on the Orasis page of your account area, are handled by the same system as Eclose Omics. Sections 3, 4, 6 and 8 describe that system.

Eclose AI (the chat and Eclose Folio)

Eclose AI is the AI chat in your account area and the AI features of Eclose Folio, our app for organising your files and lab books on your computer. The chat is available now. Eclose Folio is not available yet. This part describes what it sends once it is. Folio is planned for macOS 13.5 or later first, and for Windows later.

A request to the AI goes to account.eclose.bio, which runs on Cloudflare. From there it goes to the Eclose console, the Eclose chat service and Eclose's AI relay. They all run on our server in Amsterdam. The relay sends the request to an AI provider in the EU. The primary provider processes it in Germany. If it fails, the relay automatically uses a fallback provider, which processes it in France. Both are our processors. The request asks them not to store it and not to train on it. The answer comes back the same way.

We do not sell your data, we show no advertising, and we do not train models on your data. Eclose Folio sends details of your files, never whole files, except the text of the lab-book entry you work on and short excerpts of notes (section 3).

3. What data we process and where it comes from

For each activity: what data we process and where it comes from. For Eclose Omics, the data come from you (directly or through your AI assistant) or are created by our system when you use it. For PubVia and Orasis, they come from you and from the app on your computer. For Eclose AI, they come from you, through the chat or Eclose Folio, and from Eclose AI itself: usage counts, request logs and console audit lines.

Eclose Omics

Account and access

Your account: the user id we gave you, your role, the date we created it, and your quota. During the alpha, the quota is a fixed amount of free compute credit. We also keep a one-way hash (SHA-256) of your access token. We never store the token itself. We show it to you once, when we issue it, and keep only its hash on our server. The records of your AI-app (OAuth) connections also hold only hashes. If we approve a PubVia request for you, your account record also holds the id of your PubVia invite. The id names the invite, not you. We delete it when your account is erased, and delete the invite itself at the same moment. The install code for the invite is emailed to you and kept nowhere else. If we approve an Orasis request for you, your account record also holds the id of your Orasis download key. The id names the key, not you. We delete it when we replace the key or erase your account. If you ask for a download link for the science package, your account records when you got the last one. That tells us when you may have the next. We never store the link.

Source Us, when we create your account (user id, role, quota); our system, when it issues your token (shown to you once).

Sequencing files (FASTQ)

The sequencing files (FASTQ) you upload or ask us to download.

Source You: uploaded from your computer, or downloaded by our analysis server from links you give us (for example ENA, Dropbox or your sequencing provider).

Results, run files and job records

Your results and run files: tables, reports, quality control, the record of what was run and progress messages. Also the job record: the sample names, file names and conditions you chose, the cost of the job and any refund.

Source Created by the analysis from your files and from the settings, sample names, file names and conditions you chose. Our system calculates the cost of each job and counts it against your limits.

Security log

A security log: the time, your user id, the action, whether it worked, and a short note. Requesting a science-package download link is one of these actions. The log records that a link was issued, or that it was refused and why. It never records the link or its signature. Some entries also record your IP address. These are: connecting an AI app to your account (OAuth), using the public forms (an access request and its confirmation, or a request for a new token), and some requests refused for going over a limit. We never record your email address as the person who did something. Free-text fields hold what was typed in them. For example, the shortened error text of a failed request, or a note our staff write on a refund. The log does hold the email addresses of our staff who use the admin dashboard. See section 13.

Source Created by our system when you or your AI assistant use it. Where an entry records an IP address, it is that of your connection, as forwarded by Cloudflare.

Security monitoring: traffic counts

The traffic data we read from Cloudflare, our network provider: the IP addresses and the paths of requests. When something looks abnormal, we receive an alert email with those IP addresses and paths.

Source Our network provider (Cloudflare), which sees the requests that reach us.

Security monitoring: Cloudflare Access request log

The log of the requests to the areas Cloudflare Access protects for us, such as the admin area. It holds one entry per request, and the refused attempts are recorded too, including those of people who are not our staff. Each entry holds the email address, the IP address, the country, the app, the action, and whether access was allowed. The findings we have already raised about that log are kept for the same period, and they repeat an address.

Source Cloudflare Access, which controls who can enter our admin area.

Rate limiting

Your IP address and how many requests you made, held in our server's memory.

Source Your connection (the IP address, as forwarded by Cloudflare) and your requests.

Sign-in on the /results page

Your token, held in the session storage of your browser tab while you use the /results page.

Source You: the token you type into the sign-in form on the /results page.

Sign-in on the account area

The 10-minute code we email to the address on your account, and the session it opens: a cookie in your browser and a record of the session on our server.

Source You: the email address on your account, when you sign in to the account area.

Feedback and requests

What you choose to send through your AI assistant's feedback tool: a request, suggestion, problem report or opinion. Also what you write in the “Report a problem” box on the Orasis page. We store your user id, the text, the job it is about if you name one, the product it is about if you send it from a page, and the time. The text is free text, and it may contain personal data. A message through your assistant is sent only if you agree, and in your own words. The box is yours to fill in or leave empty. We store no IP address and no token with it.

Source You, through your AI assistant's feedback tool or the “Report a problem” box on the Orasis page of your account area.

Recovery email (optional)

Your recovery email address, if you choose to add one. Also an address you have entered but not yet confirmed. Also the one-time links we email to it, stored only as hashes. When you connect an AI app, you can ask for a 6-digit code. We keep the code only in memory, as a hash, until you use it, it expires, or you ask for a new one. It stops working as soon as the address is changed or removed. A 6-digit code that approves a change of your recovery email works the same way. It goes to your current address, never to the new one. If you turn on job-finished emails, each one carries the analysis id, whether it finished, the time, and two links: your results page, and the page where you turn these emails off. Nothing else about the analysis is in it: no file names, sample names, errors or costs. One more email goes to the same address without asking. If a run's results are kept for more than 45 days after the run finishes, we email you up to 7 days before they are deleted. That email carries the run's name, its id, the day, and the address of the run's own page, and nothing else. It goes only to an address you confirmed yourself. We never send it to an address we entered on your behalf. We share this address only with the email provider that sends the message.

Source You, on your account page; or us, if you asked us to set the address for you.

Access request (the public form)

What you write in the form: your name (up to 100 characters), your institution or lab if you give one (up to 150), your operating system, and the products you ask for. Also your answer to “What will you use it for?” (up to 1000 characters of free text, which may contain personal data). We also keep the time you ticked the privacy box, the version of the Terms of Use you accepted, and your email address. We store the one-time confirmation link only as a hash. We do not store your IP address with the request. Our rate limits hold your connection's address in memory only. They allow 5 requests an hour and 20 a day per connection, 3 a day per email address (kept as a digest), and 30 confirmations an hour. Each attempt to send the form is recorded in the security log, with your IP address. Section 13 says what the security log holds, and how long we keep it.

Source You: the public form at /request-access, after you confirm your email address by clicking the link we send you.

Access requests and correspondence by email

Your email address and what you write, when you email us to ask for access or about anything in this notice, and our reply.

Source You: the email you send to the contact address.

Passkeys

For each passkey on your account, we keep its credential id, its public key, and the counter its device keeps (the signature counter). We also keep the transports it answered with, the name you gave it, the day you added it, and the day it was last used. We also keep one flag per account: whether we have already offered you a passkey. We offer it once, and the flag records that we did. The key itself never leaves your device. What we keep is the public half, which cannot sign anything. No biometric data ever reaches us: no fingerprint, no face, no PIN. Your device checks them, and tells us only that it was you.

Source You: on your account page, when you add a passkey to your account.

PubVia

Text and pictures you send to your AI provider

When you use an AI feature, PubVia sends what that feature needs straight from your computer to the AI provider you set up. The AI features are: the chat, ghost text, the rewrite command (Cmd-K), a reply to a comment, a paper summary, and the Checks judge. What is sent is the relevant part of your manuscript. For example, for ghost text it is up to 4,000 characters before the cursor and 600 after. For a summary it is up to 24,000 characters. It also includes your memory notes. For the chat, it is your message, the conversation so far, and the results of the tools the chat used. Those results can include the text of an open-access paper the chat read. You choose the provider. It can be Anthropic or OpenAI with your own API key, ChatGPT with your own sign-in, or another OpenAI-compatible service at an address you enter. If you sign in with ChatGPT, PubVia also sends OpenAI a random identifier it created for this installation. That goes only to OpenAI. With an OpenAI API key, PubVia asks OpenAI not to store its answers (the API's “store: false” option). During a chat turn, the provider may also search the web, up to three times. From PubVia 0.1.58, when you open a figure, PubVia sends a small picture of each panel (about 640 pixels wide) to your AI provider, to recognise what each panel shows. It does this on its own, with no button, whenever an AI provider is set up: Anthropic or OpenAI with an API key or a ChatGPT sign-in, or another OpenAI-compatible service. PubVia keeps the answer next to the figure, so the same picture is not sent again to the same model. From PubVia 0.1.60, the “Find a source online” button in Checks also asks your chat model which paper a sentence cites. It sends the title of your manuscript (up to 200 characters), the sentence, and the two sentences before it (up to 400 characters each). None of this passes through us.

Source You: what you write, the figures you add and the provider you choose.

Online search for sources

When no source in your library supports a sentence you write, the Papers pane looks for one online. It does this automatically, unless you turn off Settings > “Search online automatically”. That setting is on by default. The “Find a source online” and “Search online” buttons start the same search by hand. What the search sends depends on your version of PubVia. From PubVia 0.1.50, the search sends the whole sentence to OpenAlex through our server at pubvia.eclose.bio. Its spacing is tidied first. A sentence longer than 400 characters is not sent. If our server refuses the request, PubVia sends the sentence straight to OpenAlex instead. Either way, the search also sends up to 12 key terms from the sentence straight from your computer to Europe PMC. It does not send the sentence itself. Crossref is asked last, straight from your computer. The automatic search gives Crossref the same key terms, and only when OpenAlex does not answer and Europe PMC finds nothing. A button press gives Crossref the whole sentence, up to 400 characters, and only when the other sources find nothing. When you press a button, the sentence also goes to our server. Our server passes only that text on to Valyu, a scholarly search service that searches PubMed, bioRxiv and medRxiv. From 0.1.50, the automatic search never goes to Valyu. When the chat searches for papers, it runs the same search as a button press. It uses search words that your AI provider writes. They may include words from your document or the chat. A query longer than 400 characters is not sent. In earlier versions, the automatic search and the buttons work the same way. PubVia sends the whole sentence to OpenAlex through our server, or straight to OpenAlex if our server refuses the request. It also sends the sentence to our server, which passes that text on to Valyu. Only when those find nothing does PubVia send the sentence straight from your computer to Crossref. These versions do not search Europe PMC. The chat's paper search works the same way, with the search words your AI provider writes. In every version, citation suggestions send up to 12 key terms from a sentence to OpenAlex through the same server, or straight to OpenAlex if our server refuses the request. The sentence itself is not sent. From 0.1.50, the same key terms also go straight to Europe PMC. Only when OpenAlex does not answer, and, from 0.1.50, Europe PMC finds nothing, do they go straight from your computer to Crossref. Citation suggestions never reach Valyu. Our server passes the sentence or the terms on without your IP address or the app's details. From PubVia 0.1.60, when your chat model names a paper for “Find a source online”, PubVia looks up the title, first author and year it gave, on Crossref. It does this straight from your computer. A search you start, by a button or in the chat, also sends the research field you choose in Settings > General (Biology unless you change it) to our server. Our server uses it to choose which collections Valyu searches. Valyu receives that list of collections, not the field. The automatic search and citation suggestions do not send the field.

Source You: the sentences you write.

Bibliographic checks

To check the references in your library, PubVia sends the identifiers and titles of its papers straight from your computer to Crossref, OpenAlex, Europe PMC, PubMed (NCBI), DataCite and arXiv. The identifiers are the DOI, the PubMed id and the arXiv id, plus words from a title. Never your own text. It does this when a library opens, once a day, and when you import papers. When the chat reads an open-access paper, PubVia fetches it by its DOI from these services. The paper's text then goes to your AI provider as part of the chat.

Source Your library: the references you collect.

Error reports

Unless you turn off Settings > “Send error reports” (on by default), PubVia sends us a report about 15 seconds after it starts. It also sends one whenever something fails without telling you. A report holds the app version, the operating system and processor type, the day, fixed codes for what failed and where in the app's code, and how often it happened. Except after a crash, it also holds the last part of the app's technical log, up to 4 MiB. Before sending, PubVia removes paths, file names, web addresses, email addresses, IP addresses, and user and computer names from the log. A report holds no account, no IP address and no identifier of your installation. Other free text in the log stays, and it may contain words from your work, including names. Turning the setting off also deletes any reports still waiting to be sent.

Source The app on your computer.

Update check

About 30 seconds after PubVia starts, then every 6 hours, PubVia asks our server at pubvia.eclose.bio whether a new version exists. It also asks when you check from Settings > About. The request carries the app version and a key shared by every copy of that version. Like every request, it reaches our server with your IP address. The server does not store that address. You cannot turn the check off. It is how fixes reach every copy of the app.

Source The app on your computer.

Orasis

Your download link by email

When we approve your Orasis request, or you ask for a new link in your account area, we email you a personal download link. We send it through our email provider (section 6). The email holds your address and the link.

Source You: your access request, or your request for a new link.

Update check

About 30 seconds after Orasis starts, and then every 6 hours, Orasis asks our server at orasis.eclose.bio whether a new version exists. If there is one, it downloads the signed installer from there. The request carries the app version and a key that every copy of Orasis shares. Nothing in it identifies you or your computer. Like every request, it reaches our server with your IP address. Our server does not store that address, and keeps no record of the check. You can turn the automatic check off in Settings (“Automatic update checks”, on by default); checking by hand still works. An update downloaded during the current session still installs when you quit Orasis.

Source The app on your computer.

Problem reports

Orasis sends us a problem report only if you have turned on report upload in Settings. It is off when you install the app. Even then, the report goes only when you press “Send” on it, after Orasis has shown you the whole report. Nothing is sent automatically, and nothing is retried. A report has two parts. The first is a set of fixed fields: the app version, operating system and processor type, the day (UTC), the kind of problem, the build, the tool name, the error code, and the names (not the values) of the tool's arguments. It also holds a short reply text, how long it took, the Fiji and Python versions, a log id, and facts about the image involved: its format, size class, series, channels, z and time points, bit depth and pixel size. The second part is the last part of Orasis's server log, up to 1 MiB. A report holds no image data, no file names or paths, and no text from your AI conversation. Orasis refuses to send a report if any field looks like a path, a drive, an image file name, or a name you gave your own data. Those names include image folders, lab-book titles, batch labels and macro titles. The log is cleaned by pattern, so a name could still slip through in its free text. Reports you have not sent wait on your computer: at most 50, for at most 30 days.

Source The app on your computer, when you send a report.

Downloads with your personal link

To install Orasis you use a personal download link we email you. The link carries a key we made for you. Our download service keeps a record of that key: its id, a label, when it was made, how many times it was used, and the time and version of the last download. The label is your account id, or your access request's id for your first link. The service also records each download: the key id, the time and the version. It never records your IP address. Because we made the key for you, we can tell that these records are yours.

Source Your download, with the link we sent you.

What your AI client receives

Orasis has no AI of its own, and it calls no AI provider. It runs a local MCP server on your computer (127.0.0.1). Any MCP client can use it, for example Claude Desktop or Claude Code. What Orasis returns to your client goes on to your AI provider. By default this includes the names of your image files and folders, your lab-book and saved-macro titles, and numbers and measurements. Image pixels, in overlays or previews, are sent only if you choose that in Settings. Image sharing is off when you install the app. Code your AI writes for a detector runs in a Windows sandbox with no network access, and it cannot read your other files. If the sandbox is not available, the code runs only if you turn on “Allow AI code without protection”, which is off by default. Then it runs with your own access to your files and the network, and Orasis marks the runs affected. You can turn AI-written code off in Settings → Connections → Privacy (“Turn off”); it is on by default. With it off, any step that runs AI-written code is refused. None of this passes through us.

Source You: your images, files and lab book, and what you ask your AI client to do.

Files you download by link

When you or your AI client ask Orasis to download a file from a web address, Orasis fetches it straight from that address. It uses https, and only public addresses. The site sees your IP address and the request. Google Drive share links are fetched from drive.usercontent.google.com. This does not pass through us.

Source You or your AI client: the web address.

Eclose AI (the chat and Eclose Folio)

Chat messages and answers

When you use the chat in your account area, we send your messages and the conversation so far, up to the last 40 messages, to the AI provider with your request, and we return its answer. We also send only the remembered facts in scope for that request (next item). If you turn on “Let the chat read your Omics analyses” on your account page, the chat may read your own Omics runs and results through the Omics tools. What it reads is sent to the AI provider with your request. It only reads. It never starts or changes a run. That switch is off by default.

Source You: what you write in the chat. With the Omics switch on, your Omics runs and results.

Remembered facts

The chat can remember facts about your account. Eclose Folio's overview, plan and lab-book tasks may suggest facts, but they save none. A suggested fact is saved only when the AI app you connected to Folio calls organizer_remember. Folio's help never suggests facts. The chat sends only the remembered facts in scope for your request.

Source You, through what you write in the chat or send from Eclose Folio.

Eclose Folio: the overview and plan of your files

Eclose Folio is not available yet. This part describes what it sends once it is. The overview sends a summary for each folder: the path, the counts, the date range, the categories, the instruments, the bundle kinds, three example names and a hint from the files there. It also sends the folders already in your library, your answers to its questions, and short text notes: only .txt and .md files up to 256 KB that are not personal or hidden, the first 4 KB of each, up to 64 per request. The plan sends the per-file rows: the path it shows, the name, size, category, date modified or acquired (year-month-day), instrument, software, up to 64 channel names, a duplicate mark and a local hint. For up to 50 items the AI asks to look at, it also sends the acquired date as an ISO timestamp, calibration, the members of a bundle, up to 512 channel names and the text of a note. A plan goes in batches of 200 items, up to 12,000 items. The setting “Show full file paths to AI” is on by default. While it is on, each file's full path is sent, and it can contain your own name, for example /Users/<name>/. When it is off, paths are relative to the folders you picked. Folio never sends whole files, except those note excerpts. It never sends image pixels or sequence data. Of a table file, it uses only a fixed list of header fields.

Source Your computer: the files and folders you picked, and your answers.

Eclose Folio: lab books

Eclose Folio is not available yet. This part describes what it sends once it is. When you use “Talk to the AI” in a lab book, Eclose Folio sends the entry's date, title and text (the last 20,000 characters), the experiment folder's name, the type, the field values, and the conversation, up to 20 messages of 2,000 characters each. When you use “Draft from the files”, it also sends up to 400 file rows of that experiment and up to 32 notes of 4 KB each. A file row holds the path relative to the experiment, the size, the dates, the category, the instrument, the software and the channels.

Source You: your lab book and the files of the experiment.

Eclose Folio: help

Eclose Folio is not available yet. This part describes what it sends once it is. When you ask the help a question, Eclose Folio sends the question, up to 4,000 characters, and what is on your screen: the step, up to 300 visible row names, the buttons, the questions and up to 10 messages. It sends no remembered facts. It never sends file contents or full paths. The help never saves a remembered fact.

Source You: your question and what is on your screen.

Eclose Folio through your own AI

Eclose Folio runs a local MCP server on your computer (127.0.0.1), with a key that stays on your computer. An AI app you connect, for example Claude, reads the same details of your files that Folio sends for its own tasks, and it can move files. A move can be undone. It can also list and read your lab-book entries, with the full text of an entry, create an entry, append to one, replace one, and read what is on your screen. This traffic goes from your computer to the AI provider you chose. It does not pass through us. Three tools are the exception: organizer_ai_plan, organizer_remember and organizer_forget. They call Eclose AI as described in the items above.

Source You: your files, and what you ask your AI app to do.

Eclose Folio sign-in

Eclose Folio signs in with your Eclose account. Our account service keeps a record of the grant. The access token is valid for 1 hour and the refresh token for 30 days. The app keeps only the refresh token, in your operating system's credential store.

Source You: your sign-in to your Eclose account.

Usage counts

For each account and product, we count the tokens the AI used. A token is a small piece of text. The relay keeps the counts. Our console shows only what the relay sends.

Source Eclose AI: it counts your requests and the answers.

Request logs

The chat service writes one request line with the method, the route pattern, not the raw path, and the status. It writes separate lines with the name of a tool and how long it took, and with token counts. The lines hold no message text and no account id. The relay also writes one line per request.

Source Eclose AI: the request.

Console audit lines

Our server writes one line in its security log for each call to the chat. The line holds your user id, the route pattern of the call, its status and the time. It holds no message text.

Source Eclose AI: the call to the chat.

4. Why we process data and on what legal basis

Eclose Omics

Eclose Omics: what we do, why, on what legal basis and for how long
PurposeDataLegal basisRetention
To recognise you and apply your limits.Account and accessArt. 6(1)(b) — to provide the service you asked forUntil we delete your account; an AI-app connection, 30 days after its last use
To run the analysis you ask for.Sequencing files (FASTQ)Art. 6(1)(b) — to provide the service you asked for4 days (the storage deletes them after 30 days if they are still there)
To deliver the analysis and let you check how it was produced.Results, run files and job recordsArt. 6(1)(b) — to provide the service you asked forResults: 365 days from the day the run starts for a run started now, or 30 days for a run started before (the date is shown on each run). Job record on our server: at least 90 days, then reduced to a billing summary.
To detect abuse and settle questions about who did what.Security logArt. 6(1)(f) — legitimate interest365 days (12 months); IP addresses 90 days
To keep the service secure.Security monitoring: traffic countsArt. 6(1)(f) — legitimate interest8 days
To keep the service secure.Security monitoring: Cloudflare Access request logArt. 6(1)(f) — legitimate interest12 months
To stop anyone flooding the service.Rate limitingArt. 6(1)(f) — legitimate interestMemory only: one minute (up to 24 hours for the recovery-email limits); lost on restart
To keep you signed in while the tab is open.Sign-in on the /results pageArt. 6(1)(b) — to provide the service you asked for; the browser storage is strictly necessary for the page you asked forUntil you sign out or close the tab
To recognise you in the account area without an AI app, so you can use the service yourself.Sign-in on the account areaArt. 6(1)(b) — to provide the service you asked for; the cookie is strictly necessary for the session you asked forThe code: 10 minutes. The session: until you sign out, or 14 days
To improve the service and answer you.Feedback and requestsArt. 6(1)(f) — legitimate interest (improving the service and answering you)6 months
We use your recovery email only to:
  1. get a new token if you lose yours, or receive an invite if we send you one
  2. sign in to the page where you connect an AI app, with a 6-digit code emailed to you instead of a link
  3. approve a change of this address: a 6-digit code goes to your current address, and a notice with no link or code goes to the address being replaced
  4. tell you that the copy of your data you asked for is ready to download (the email carries no data)
  5. warn you, at the address you confirmed yourself, up to 7 days before one of your runs' results is deleted, if they are kept more than 45 days after the run finishes
  6. tell you when one of your analyses has finished, only if you turn on job-finished emails
Uses 1 to 5 always apply. Use 6 applies only if you turn it on.
Recovery email (optional)Art. 6(1)(a) — consent (an invite: Art. 6(1)(b))Until you remove it. Unused links: 30 minutes, 24 hours or 7 days, by kind. A sign-in code: 10 minutes. The ids of jobs we have already announced: 90 days
So we can read and decide on your request by hand and, if we grant access, create your account and send you what you need.Access request (the public form)Art. 6(1)(a) — consent (the privacy box and the confirmation click are the proof)Unconfirmed: 24 hours. Confirmed: until every product is decided, then at most 12 months
To consider and answer your request and, if we invite you, to create your account.Access requests and correspondence by emailArt. 6(1)(b) — steps you ask us to take before an agreement; otherwise Art. 6(1)(f) — legitimate interest in answering messagesUntil the request is closed, then at most 12 months — unless we create an account, when the account's rules apply
To let you sign in with your device (a fingerprint, face or PIN) instead of an emailed code. The emailed code keeps working, so a passkey is an addition and never a replacement.PasskeysArt. 6(1)(b) — to provide the service you asked forUntil you remove the passkey or your account is erased

More about the legal bases

PubVia

PubVia: what we do, why, on what legal basis and for how long
PurposeDataLegal basisRetention
The AI feature you use.Text and pictures you send to your AI providerNone from us: we do not receive this data (section 6)Under your AI provider's terms
Finding published sources for what you write.Online search for sourcesArt. 6(1)(f) — legitimate interestNot kept by us
Checking that the references in your library exist and are described correctly.Bibliographic checksArt. 6(1)(f) — legitimate interestNot sent to us
Finding and fixing faults in the app.Error reportsArt. 6(1)(f) — legitimate interest30 days
Telling you that a new version is available.Update checkArt. 6(1)(f) — legitimate interestNot stored

Orasis

Orasis: what we do, why, on what legal basis and for how long
PurposeDataLegal basisRetention
Sending you Orasis, which you asked for.Your download link by emailArt. 6(1)(b) — to provide what you asked forWe keep no copy of the email
Getting fixes and new versions to you.Update checkArt. 6(1)(f) — legitimate interestNot stored
Finding and fixing faults in the app.Problem reportsArt. 6(1)(a) — consent (turning on report upload, and pressing “Send” for each report)30 days
Seeing whether and when you downloaded Orasis, and which version, so we can support you and notice a link that is being misused.Downloads with your personal linkArt. 6(1)(f) — legitimate interestEach download: 90 days. The key's record: until we replace the key or erase your account
The analysis you ask your AI client to run.What your AI client receivesNone from us: we do not receive these data (section 6)Under your AI provider's terms
Getting the file you asked for.Files you download by linkNone from us: we do not receive these dataNot sent to us

Eclose AI (the chat and Eclose Folio)

Eclose AI (the chat and Eclose Folio): what we do, why, on what legal basis and for how long
PurposeDataLegal basisRetention
Answering what you ask the chat.Chat messages and answersArt. 6(1)(b) — performance of a contractConversations: 7 days after the last message
Giving the AI what it learned from you before.Remembered factsArt. 6(1)(b) — performance of a contract12 months after it was saved or last edited
Organising your files with the AI, which you asked for.Eclose Folio: the overview and plan of your filesArt. 6(1)(b) — performance of a contractAnswers: 10 minutes, in memory only
Working on your lab book with the AI, which you asked for.Eclose Folio: lab booksArt. 6(1)(b) — performance of a contractAnswers: 10 minutes, in memory only
Answering your question about Eclose Folio.Eclose Folio: helpArt. 6(1)(b) — performance of a contractAnswers: 10 minutes, in memory only
The work you ask your AI app to do.Eclose Folio through your own AINone from us for what your provider receives (section 6). For the three tools: Art. 6(1)(b) — performance of a contractUnder your AI provider's terms
Letting Eclose Folio use your account.Eclose Folio sign-inArt. 6(1)(b) — performance of a contractAccess token 1 hour; refresh token 30 days
Enforcing the usage limit, and showing you and our staff the usage.Usage countsArt. 6(1)(b) — performance of a contract7 days for the limit, 90 days in our admin counters, 9 weeks in your usage view
Running the chat and the relay, and finding faults.Request logsArt. 6(1)(f) — legitimate interest30 days
To detect abuse and settle questions about who did what.Console audit linesArt. 6(1)(f) — legitimate interest365 days (12 months)

Legitimate interest and your right to object

Eclose Omics. We rely on legitimate interest for four things. (1) Keeping the service secure and preventing abuse: the security log and rate limiting. (2) Protecting our cloud spending, and settling disputes over quotas and refunds: the security log. (3) Improving the service and answering you: feedback. (4) Answering the messages we receive: emails. We have weighed these interests against yours. We concluded that they do not override your rights. The data are limited to what is needed. An IP address goes to disk only in the security log. That is for connecting an AI app, the public forms, and some refused requests. It is replaced after the period in section 8. Feedback carries no IP address and no token, and it is sent only if you agree. Every item has a fixed, short retention period. We do not use the data for profiling or for advertising. We do not use them for any other purpose. We share them with nobody except the recipients in section 6.

PubVia. We rely on legitimate interest for the online search, the bibliographic checks, the error reports and the update check. They are how the app finds sources, checks references, gets its faults fixed, and gets new versions to you. We have weighed these interests against yours. We concluded that they do not override your rights. Each sends only what its purpose needs, and none of it is tied to an account. We keep none of it for more than 30 days, and most of it not at all. You can turn off the online search and the error reports in Settings at any time.

Orasis. We rely on legitimate interest for the update check and the download records. They are how fixes reach you, and how we support testers and notice a misused link. We have weighed these interests against yours. We concluded that they do not override your rights. The update check carries nothing that identifies you, and our server stores nothing about it. The download records hold no IP address, and each download is deleted after 90 days. You can turn the automatic update check off in Settings. Problem reports rest on your consent instead.

Eclose AI (the chat and Eclose Folio). We rely on legitimate interest for the request logs and the audit lines of our server. They are how we run the chat and find faults, and who did what. We have weighed this interest against yours. We concluded that it does not override your rights. A request line holds no message text or account id, and we delete it after 30 days. An audit line holds your user id, and we delete it after 365 days.

You can object at any time to processing that rests on a legitimate interest. Section 9 explains how.

5. Sensitive data

Eclose Omics

Eclose Omics is meant for sequencing data from cell lines and other research material that does not identify a person. Do not upload data about an identified or identifiable person, and do not put such data in sample names, file names, job names or feedback.

We accept human samples only from cell lines. You confirm this when you submit them. We do not accept patient or clinical material. We do not accept controlled-access data, even from cell lines. Examples are data from dbGaP or EGA, and data from any repository that requires a data access agreement. Their terms do not allow processing on this service.

Our system enforces this for human data: it refuses a human job unless the submission carries the “cell-lines-only” confirmation. We do not intend to process special categories of personal data (Art. 9 GDPR), such as a person's genetic or health data, and the service is not designed for them. If you find that such data were submitted, tell us and we will delete them.

PubVia

What you write may contain any kind of data. Do not put data about an identifiable person, such as a patient, in a sentence you write while automatic online search is on. The same applies to text you send to your AI provider. Do it only if you are allowed to share that data with those services. Turn off error reports while you work on such text. Their log can carry words from it.

Orasis

Your images may show anything, and Orasis keeps them on your computer. Your AI client receives the names of your files and folders and your lab-book and macro titles by default, so do not put a person's name or other identifying data in them. Leave image sharing off for images of identifiable people, such as patient material.

Eclose AI (the chat and Eclose Folio)

What you write to the chat, and the file names, paths and notes Eclose Folio sends, may contain any kind of data. Send data about an identifiable person, such as a patient, only if you are allowed to share it with the AI providers in section 6. The full file paths that Eclose Folio sends by default can contain your own name. Turn off “Show full file paths to AI” to send paths relative to the folders you picked.

6. Who receives your data

Eclose Omics

We store your files and results in the EU, with Scaleway, in region fr-par (Paris, France (EU)).

We keep everything else we hold on our server (a Scaleway machine in Amsterdam, Netherlands (EU) that we run ourselves). That includes account records with token hashes, job records, the security log, the sign-in log of the admin area, feedback, recovery email addresses, OAuth records, and the configuration. It also includes the onboarding register: our own staff records, a colleague's name, work email address and the day they confirmed our rules. The register is kept for three years after that person's access ends, and gone from every remaining backup at the latest 30 days after that. Section 8 lists each item.

We have access to all the data described in this notice. Our staff have the access their role gives them in the admin dashboard. Staff with the admin role can read users' job records, balances, money movements, feedback, access requests and the security log. They can decide access requests, cancel or refund jobs, and change a user's quotas. Staff with the viewer role have a reduced, read-only view. Our processors and other recipients are:

Scaleway SAS

Hosts this service in the EU. It hosts our server, a Scaleway machine in Amsterdam, Netherlands (EU) (zone nl-ams-1) that we run ourselves. It also hosts the computing and object storage in region fr-par. Your FASTQs, results and job records are kept there, and the virtual machines that run your analysis are created there.

Where Amsterdam, Netherlands (EU) (zone nl-ams-1) for our server; Paris, France (EU) (region fr-par) for analysis machines and object storage.

Processor under its General Conditions of Services and Data Processing Agreement.

Scaleway contracts (General Conditions and DPA) · Scaleway privacy policy

Cloudflare, Inc.

Runs the encrypted tunnel that carries requests to our server. It also stores the offsite copy of our nightly backup. Cloudflare decrypts the connection at its edge for console.eclose.bio, then forwards each request to our server through the tunnel. So it can technically see what passes through. That includes access tokens in headers, OAuth tokens, MCP tool arguments and results, job details, signed links, and the /results and consent pages. In its storage, it keeps only one thing: the encrypted archive our server writes each night. See the backup row in section 8.

Where Global network for the tunnel. The offsite backup bucket is in the European Union. EU transfer safeguards under its Customer DPA.

FASTQ and result files do not pass through Cloudflare. Uploads and downloads go directly between your computer and Scaleway object storage, through signed links. The offsite backup bucket holds the same encrypted archive we write to Scaleway Object Storage. It is encrypted before upload, with the key kept by us. It holds nothing else: no request, no request log and no address. Only our server can write to it. Cloudflare acts as our processor under its Customer DPA. It keeps technical logs under its own terms.

Cloudflare Access also signs our staff in to the admin dashboard at admin.eclose.bio. It checks each staff member's email address. That traffic goes through the same tunnel. So Cloudflare can technically see what the dashboard shows, including users' data. See section 13.

Cloudflare keeps its own technical logs for the period set in its terms and privacy policy (linked here). We cannot shorten that period, and an erasure on our side does not reach those logs.

Cloudflare Customer DPA · Cloudflare privacy policy · Cloudflare Turnstile privacy addendum

Resend, Inc.

Sends our emails. Resend receives the address and the content of each message.

What we send through Resend

What Resend receives in each case

Where Our sending domain is set up in Resend's eu-west-1 (Ireland) region, which decides only where emails are sent from. Resend stores account data, email metadata and logs in the United States.

Processor under Resend's Data Processing Addendum (Art. 28). It applies automatically and includes the EU Standard Contractual Clauses (Module Two) for the transfer to the United States. Resend can read the messages it sends. Anyone with access to your mailbox, or to Resend's copy, could open a link in them. Each link works once, for a limited time. It shows a token only after you press a button.

Resend keeps its own technical logs for the period set in its terms and privacy policy (linked here). We cannot shorten that period, and an erasure on our side does not reach those logs.

Resend Data Processing Addendum · Resend privacy policy

Emails you send to the contact address are stored, on our behalf, by the provider that hosts that mailbox.

Your AI provider

Your AI provider is not our processor. For what your assistant receives, it is an independent controller, under the agreement you made with it.

The provider of the AI assistant you connect (for example Anthropic for Claude, or OpenAI for Codex) receives everything our server returns to your assistant. That includes the job status and details, the sample names, the warnings, and the result summaries (QC, top genes, tables your assistant asks for). It also includes the Methods text, the download links, and the pipeline code you ask for. If you connect through a connector in a web or desktop app (for example claude.ai or Claude Desktop), the requests to our server come from the provider's servers. Those servers also hold the access token for your session. If your assistant downloads and reads result files, their content reaches the provider too. This happens under your own agreement with your AI provider, not under our control. Check your provider's data and training settings before you work with sensitive data. Files you download yourself, from the links or the /results page, go only between your computer and Scaleway.

We give your AI assistant download links for your result files only during the first 24 hours after a job finishes. Each link expires by the end of that window. After that, the assistant gets the list of files without links. You download them from your results page (/results, signed in with your token). That page keeps providing links for the whole retention period.

FASTQ files are not sent to your AI provider. Uploads go straight from your computer to Scaleway through the upload link. Files you give us as links (ENA, Dropbox, your sequencing provider) are downloaded by our analysis server. Only if your assistant opens a file to look inside it, for example at its first reads, does that excerpt reach your provider.

Accession lookups: when your assistant asks us to look up public ENA or GEO accession numbers, our server queries ENA's public file report and NCBI GEO's accession page itself. Each request carries only the public accession number. It has no credential and no user identifier. Those services see our server's IP address, not yours.

PubVia

PubVia runs on your computer. Our server for PubVia, and the services the app contacts, are:

Cloudflare, Inc.

Runs our server for PubVia (pubvia.eclose.bio: online search, error reports, updates and downloads). It also stores the error reports. It sees every request that reaches pubvia.eclose.bio: your IP address, the app's details, the address requested, and the content. The content is the sentences and terms of an online search, the research field of a search you start, and the error reports.

Where Global network for the server; error reports are stored in the EU (an R2 bucket under EU jurisdiction).

Processor under its Customer DPA. We have turned off request logging on these services. Cloudflare keeps its own technical logs under its terms, which we cannot shorten.

Cloudflare Customer DPA · Cloudflare privacy policy

Valyu (Valyu.Network LTD)

A scholarly search service that searches PubMed, bioRxiv and medRxiv. From PubVia 0.1.50, through our server and with our key, it receives only the text of a search you start with a button or the chat starts. It never receives your IP address or an identifier of the app. The automatic search never reaches it. In earlier versions, the Papers pane's automatic search, its search button, Checks' online search and the chat's paper search all reach it. Citation suggestions do not. From 0.1.60, it also receives the list of its collections that our server chose from your research field.

Where The United Kingdom, the European Union and, for some of its infrastructure vendors, the United States (Valyu's own statement).

Valyu says that it does not use query data or logs to train machine-learning models. It says that it deletes query content automatically within a short window (typically 24 to 72 hours), and that it keeps security logs for up to 90 days. These are Valyu's published terms. We have signed no separate agreement with it.

Valyu privacy policy

OpenAlex (OurResearch)

A public index of research papers. Through our server, with our key and without your IP address, it receives the sentences and key terms of the online search. Straight from your computer, with your IP address, it receives the search when our server refuses it. It also receives the identifiers and titles of the bibliographic checks.

Where The United States.

OpenAlex says it keeps a log of each API request for 90 days. The log holds the search terms, the IP address and the app's details, linked to the API key.

OpenAlex privacy policy

Crossref

A public registry of DOIs. Straight from your computer, with your IP address, it receives a search when the other sources find nothing. From PubVia 0.1.50, that is the key terms of an automatic search, or the sentence of a search you start. In earlier versions, it is the sentence of any search. It also receives the key terms of a citation suggestion, but only when OpenAlex does not answer. From 0.1.50, Europe PMC must also find nothing. It receives the identifiers and titles of the bibliographic checks too. From 0.1.60, it also receives the title, first author and year of a paper your chat model names for “Find a source online”.

Where The United States.

Under its own privacy policy (linked here).

Crossref privacy policy

Europe PMC, PubMed, DataCite and arXiv

Public bibliographic services. Straight from your computer, with your IP address, they receive the identifiers and titles of the bibliographic checks, and the DOI of any paper the chat reads. From PubVia 0.1.50, Europe PMC also receives the key terms of each online search: up to 12 words from a sentence you write. None of them receives a whole sentence.

Where Europe PMC: United Kingdom (EMBL-EBI). PubMed: United States (NCBI). DataCite: Germany. arXiv: United States (Cornell University).

Each acts under its own terms (linked here).

Europe PMC privacy notice · NLM web policies (PubMed) · DataCite privacy policy · arXiv privacy policy

Your AI provider

Your AI provider is not our processor. PubVia sends your text, and from 0.1.58 also pictures of your figures' panels, to it straight from your computer. For what it receives, the provider is an independent controller, under the agreement you made with it. That agreement decides whether it keeps your text and whether it trains its models on it.

At the date of this notice, these are the terms. Anthropic's Commercial Terms cover use with an API key. They say that “Anthropic may not train models on Customer Content from Services”. OpenAI's page on business data covers its API platform. It says that by default it does not train its models on business data. It also says that, except for certain endpoints and features listed in its documentation, it may keep API inputs and outputs for up to 30 days. That is to provide the service and to identify abuse. With ChatGPT sign-in, your own ChatGPT data controls apply. On personal ChatGPT Free, Plus and Pro workspaces, the setting “Improve the model for everyone” is on, unless you turn it off. With another service at an address you enter, that service's terms apply. We promise nothing beyond these terms.

Anthropic Commercial Terms · OpenAI: business data and the API · Data controls in ChatGPT

Orasis

Cloudflare, Inc.

Runs our server for Orasis (orasis.eclose.bio: downloads, updates and problem reports). It stores the problem reports, the installers and the download records. It sees every request to orasis.eclose.bio: your IP address, the app's details, the address requested, and the content of a problem report.

Where Global network for the server. Problem reports are stored in the EU (an R2 bucket under EU jurisdiction). Download records are stored in Eastern North America (a D1 database).

Processor under its Customer DPA. We have turned off request logging on these services. Cloudflare keeps its own technical logs under its terms, which we cannot shorten.

Cloudflare Customer DPA · Cloudflare privacy policy

Sites you download from

Any site you or your AI client ask Orasis to download a file from. It receives the request and your IP address, straight from your computer.

Where Wherever the site is.

Each acts under its own terms.

Your AI provider

Your AI provider is not our processor, and Orasis does not send it anything: your AI client does, under the agreement you made with your provider. For what it receives, the provider is an independent controller. That agreement decides whether it keeps the data and whether it trains its models on it.

Eclose AI (the chat and Eclose Folio)

The chat and Eclose Folio send their requests through our server. The services that handle a request are:

Cloudflare, Inc.

Carries your requests to account.eclose.bio, through a tunnel to our server. It sees the requests that pass through it: your IP address and the content.

Where Global network.

Processor under its Customer DPA. Cloudflare keeps its own technical logs under its terms, which we cannot shorten.

Cloudflare Customer DPA · Cloudflare privacy policy

Primary AI provider

Runs the AI model that reads your request and writes the answer. It receives what section 3 lists for the chat and for the AI features of Eclose Folio. The request asks it not to store the request and not to train on it.

Where Germany (EU).

Our processor.

Fallback AI provider

Does the same as the primary provider, but only when the primary provider fails. Our relay then sends the request to it automatically.

Where France (EU).

Our processor.

Your AI provider

Your AI provider is not our processor. In the “your own AI” route of Eclose Folio, your AI app sends what it reads straight from your computer to the provider you chose, under the agreement you made with it. For what it receives, that provider is an independent controller. That agreement decides whether it keeps the data and whether it trains its models on it. Three tools of Eclose Folio are the exception (section 3).

7. Transfers outside the EEA

We process personal data in the EU, except as shown below. Where a transfer has a safeguard, the documents linked in section 6 contain it and tell you how to get a copy; we also send you a copy on request. Where the table says “None”, there is no safeguard.

Transfers of personal data outside the European Economic Area
RecipientDestinationSafeguard
Scaleway SASNone: our server is in Amsterdam, Netherlands (EU), and analysis and storage are in Paris, France (EU) — both in the EU.Not applicable.
Cloudflare, Inc.Possible: its network is global, and it processes the requests that pass through it.The EU Standard Contractual Clauses in its Customer DPA (https://www.cloudflare.com/cloudflare-customer-dpa/).
Resend, Inc.The United States: Resend stores account data, email metadata and logs there.The EU Standard Contractual Clauses, Module Two, in its Data Processing Addendum (https://resend.com/legal/dpa).
Your AI providerDecided by your provider.None from us: the provider is an independent controller and acts under the agreement you made with it.
Your AI provider (PubVia)The United States, for Anthropic and OpenAI (ChatGPT sign-in included); for another service, wherever it runs.None from us: your computer sends the data to the provider you chose, under the agreement you made with it.
OpenAlex and Crossref (PubVia)The United States.None: public research indexes used under their published terms. A sentence that reaches OpenAlex through our server carries no IP address.
Valyu (PubVia)The United Kingdom and, for some of its infrastructure vendors, the United States.For the United Kingdom, the European Commission's adequacy decision. For the United States, none: only Valyu's published terms, with no agreement signed by us. Only the text of the search is passed on, without your IP address.
Europe PMC, PubMed and arXiv (PubVia)The United Kingdom and the United States.For the United Kingdom (Europe PMC), the European Commission's adequacy decision. For the United States (PubMed, arXiv), none: public bibliographic services your computer contacts directly, with identifiers and titles of papers and, for Europe PMC from PubVia 0.1.50, the key terms of an online search.
Cloudflare, Inc. (Orasis download records)Eastern North America: the download records are stored there.The EU Standard Contractual Clauses in its Customer DPA (https://www.cloudflare.com/cloudflare-customer-dpa/).
Your AI provider (Orasis)Decided by your provider.None from us: your AI client sends the data to your provider, under the agreement you made with it.
Sites you download from (Orasis)Wherever the site is.None from us: your computer sends the request, under that site's own terms.
Your AI provider (Eclose Folio)Decided by your provider.None from us: your AI app sends the data to your provider, under the agreement you made with it.

8. How long we keep data

Eclose Omics

Eclose Omics: how long we keep data
DataPeriodHow it is enforced
Account: user id, role, quota, token hash, and the time of your last science-package download link (never the link)Until we delete your accountOn our server; no expiry.
AI-app (OAuth) connectionAccess: 60 minutes. Renewal: 30 days after the last useEnds at once if you or we revoke it.
Sequencing files (FASTQ)4 daysA daily clean-up on our side deletes them. If it ever fails, the storage deletes them after 30 days anyway.
Results, run files and the storage copy of the job record365 days from the day the run starts for a run started now, or 30 days for a run started before (the date is shown on each run)Results are deleted on the date shown on each run, or by the storage 379 days after they were stored, whichever comes first. After the deletion, a copy that no one can open stays in storage for up to 2 days. Then it is removed too.
Job record on our server (names, files, conditions, messages)At least 90 days after the job finishes, and in any case until its results are deletedThen reduced to a billing summary (date, status, cost, refund figures). Exception: while a question about the job's analysis machine is open (it may still be billing, or be kept for reuse), the full record stays until that is resolved. We check such open questions at least once a month.
Billing summaryUntil we delete your accountKept under your user id; no sample names, file names or results.
Account balance and its movements (top-ups, charges per job, refunds)While the account exists. After it is closed, until the end of the tenth year after its last movement; then deletedWe never store the balance itself. It is the sum of the movements, so a live account keeps its whole history. The daily clean-up deletes a closed or erased account's movements all together, once every movement of that account is past its period. Erasing the account replaces its id with a one-off pseudonym and drops the notes. The amounts, dates and job references stay for the period above (money records: Art. 6(1)(c)). After an erasure we keep a one-way hash of your account name, so that the name is never given to anyone else. We keep it indefinitely, and it contains nothing else.
Security log365 days (12 months)IP addresses in it are replaced by a placeholder after 90 days. A damaged line with no readable date is not removed automatically; we remove such lines by hand when we find them. Entries with a staff member's email address follow the same 365-day rule (section 13).
IP address and request counts for rate limitingOne minute (up to 24 hours for the recovery-email limits)Memory only; lost when our server restarts.
Feedback and requests6 monthsDeleted by a daily check; sooner if we delete it or your account is erased.
Recovery email addressUntil you remove it or your account is erasedYou can remove it on your account page at any time.
Traffic counts (the security monitor)8 daysThe monitor stores counts only. Section 13 says more.
Cloudflare Access request log12 monthsOne entry per request, and the refused attempts are recorded too. Section 13 says more.
Security alert emails to us12 monthsWe delete them from our mailbox.
Passkeys (the public key, the credential id, the counter, the transports, the name you gave it and the two dates)Until you remove the passkey or your account is erasedYou can remove one at any time on your account page. No biometric data is kept: your device checks it and tells us only that it was you.
Recovery links (one-time)30 minutes (token recovery), 24 hours (confirmation), 7 days (invite)A link works once; an expired link is deleted within 1 hour, by an hourly check.
Emails to the contact addressUntil the request is closed, then at most 12 monthsWe delete them by hand; the account's rules apply if we create an account.
Technical logs of Cloudflare and ResendSet by the provider's termsWe cannot shorten them; see the privacy policies in section 6.
Token in your browser tab (/results)Until you sign out or close the tabHeld by your browser, not by us.
Encrypted backups of our server's data30 days (the offsite copy)Made nightly and encrypted before upload, with the key kept by us. Each backup is deleted 14 days after it was made, and the offsite copy after 30 days, so the period shown is the longer one. Section 13 says where the backups go and why we keep them.

The storage also deletes a run's results about 379 days after they are stored, if the date on the run has not come first. Uploads that were never completed are cancelled after 1 day. When a file is deleted or expires, an older version of it stays. That version is removed about 1 day later. So a deletion of a result, whether it is an expiry or an erasure, is complete about 1 day after it happens.

Your uploads, and the FASTQs a new run prepares, are kept apart from the results, under a rule of their own. The storage deletes them about 30 days after they are stored, if our daily clean-up has not removed them first. Uploads that were never completed are cancelled after 1 day. Uploads made earlier are kept with the results, and the results' rule applies to them.

More about each period

PubVia

PubVia: how long we keep data
DataPeriodHow it is enforced
Error reports30 daysDeleted automatically by the storage (Cloudflare R2, in the EU) 30 days after each report arrives.
Sentences and terms of an online searchNot storedPassed on to the search services; our server keeps no copy.
Rate-limit key: a keyed hash of your IP address, per kind of search and per dayAbout 3 daysOur server keeps at most about three days of them.
Update checksNot storedOur server only reads the list of versions; nothing is written.
PubVia's data on your computerUntil you delete itHeld by you, not by us (section 12).

More about each period

Orasis

Orasis: how long we keep data
DataPeriodHow it is enforced
Problem reports you sent30 daysDeleted automatically by the storage (Cloudflare R2, in the EU) 30 days after each report arrives.
Problem reports not yet sentUp to 30 daysOn your computer, at most 50, until you send or delete them.
Each download (key id, time, version)90 daysDeleted by an hourly check.
Your download key's record (id, label, count, last download)Until we replace the key or erase your accountWe delete it on our download service.
Update checksNot storedOur server writes nothing.
Orasis's data on your computerUntil you delete itHeld by you, not by us (section 12).

More about each period

Eclose AI (the chat and Eclose Folio)

Eclose AI (the chat and Eclose Folio): how long we keep data
DataPeriodHow it is enforced
Chat conversations7 days after the last messageDeleted by a sweep that runs every hour. At most 200 per account.
Remembered facts12 months after saved or last editedDeleted by a sweep. At most 200 per account. You can delete each one.
Answers to Eclose Folio's AI tasks10 minutesHeld in memory only, never written to disk.
Usage counts: tokens per account and product7 days for the limit, 90 days in our admin counters, 9 weeks in your usage viewKept by the relay. Our console shows only what the relay sends.
Request log lines of the chat service and the relay30 daysDeleted from our server's logs after 30 days. They hold no message text and no account id.
Console audit lines for chat calls365 days (12 months)Deleted from our security log 365 days after they were written. Your user id is replaced when your account is erased.
Eclose Folio sign-inAccess token 1 hour; refresh token 30 daysErased with your account.
Eclose Folio's data on your computerUntil you delete itHeld by you, not by us (section 12).

More about each period

9. Your rights

You have the following rights. Using them is free; the end of this section explains how.

Eclose Omics

Access (Art. 15)

You can see your own jobs, results and quota on the /results page (sign in with your token) or by asking your AI assistant. For the feedback you sent, the security log or anything else, ask us.

Rectification (Art. 16)

You can change your recovery email yourself on your account page. The other records are created by our system. A user id cannot be changed, because your records are filed under it: if yours is wrong, we create a new account for you and close the old one. Ask us to correct a wrong quota. A job's content is what you submitted; if it was wrong, run the job again.

Erasure (Art. 17)

Ask us. The erasure deletes the data we hold about you. Your token and your AI-app connections stop working. We delete your recovery email if you set one. We also delete the feedback you sent, your job records and your downloaded results on our server, and your uploaded files and results in storage. Your user id in the security log is replaced by a random placeholder. Any unfinished multi-part upload is cancelled. After an erasure we keep a one-way hash of your account name, so that the name is never given to anyone else. We keep it indefinitely, and it contains nothing else. Links you were given before the erasure cannot be recalled. A single-file upload link stays usable for up to 1 hour, and other links for up to 24 hours. Download links then point at deleted files. A single-file upload that still arrives through an unexpired link is removed by our daily clean-up within about 4 days. If that clean-up ever fails, the storage deletes it after 30 days anyway. Results are removed on their run's own date. If that date has not come, the storage deletes them after 379 days anyway. An older version of a deleted file stays for about 1 more day. So the erasure is complete about 1 day after we carry it out. While one of your jobs is unfinished, or a question about its analysis machine is open, our system delays the erasure. The erasure is delayed, not refused for good. We carry it out once the question is resolved. The erasure covers what our server and storage hold. We delete emails you sent to the contact address by hand. Copies held by processors follow their own retention. These are the technical logs of Cloudflare and, if email is on, of the email provider. Your AI provider keeps its own copy, under your agreement with it. Our nightly backup holds the same data, so the erasure reaches the backups too. Section 13 says how the backups are made and where they go. Each backup is deleted 14 days after it was made, and the offsite copy after 30 days. So your data are gone from every copy at the latest 30 days after the erasure. The one exception is your account balance. After the account is closed, we keep its movements in the pseudonymised form described in section 8: the amounts, the dates and the job references, without your account name and without any note. We keep them until the end of the tenth year after the last movement, because Italian accounting and tax law requires us to keep money records. Then we delete them (Art. 17(3)(b)).

Restriction of processing (Art. 18)

Ask us at the contact address, in the cases listed in Art. 18(1) — for example while you contest the accuracy of your data, or while we examine an objection. Our system has no restriction switch, so we do it by hand. We revoke your access, so that no job runs and nothing new is processed. We keep your stored data unused until the restriction is lifted or you ask for erasure. Apart from storing it, we then process the data only with your consent, for legal claims, or as the law allows. Restriction is not erasure: your stored data is kept but not used. Short-lived technical records, such as sign-in grants and links, still expire on their own. We tell you before we lift a restriction.

Data portability (Art. 20)

You can download your results and files yourself from the /results page (or through your AI assistant), in the formats the analysis produced (tables, reports, files). For any other data you gave us, ask us for a copy in a structured, commonly used, machine-readable format.

Recovery email

You can see, change or remove your recovery email yourself, on your account page, at any time. To change or remove an existing address, you need a 6-digit code. We email it to your current address. We also tell the address being replaced, so you hear about a change you did not make. Removing the address, or erasing your account, deletes the address and every unused link. You can also turn job-finished emails on and off on your account page, at any time. Turning them off withdraws the consent they rest on, and stops them at once, even if you no longer have a confirmed address.

PubVia

PubVia has no account. What we keep from it, error reports and rate-limit keys, is not linked to an account or to an identifier of your installation. So we cannot find one person's data among them (Art. 11 GDPR), unless you show us which data are yours, for example by a report's date and content. Either way, they are deleted within the periods in section 8. The data on your computer are yours to see and delete (section 12).

Orasis

Orasis has no account. A problem report is not linked to you or your computer, so we cannot single out your reports (Art. 11 GDPR) unless you tell us which they are, for example by the date and the log id. Download records are linked to your personal download link, so we can find, show and delete them when you ask. You withdraw your consent to problem reports by turning report upload off in Settings. Reports already sent are deleted after 30 days, or sooner if you ask and we can find them. The data on your computer are yours to see and delete (section 12).

Eclose AI (the chat and Eclose Folio)

Your chat conversations, remembered facts and Eclose Folio sign-in are linked to your account, so we can find, show and delete them when you ask. You can delete each remembered fact yourself. Conversations are deleted 7 days after the last message. Erasing your account erases the conversations, the remembered facts and the sign-in grants. The data on your computer are yours to see and delete (section 12).

For every product

Objection (Art. 21)

You can object at any time, on grounds relating to your particular situation, to processing based on legitimate interest (Art. 21(1) GDPR). Write to the contact address. We then stop, unless we can show compelling legitimate grounds that override your interests, rights and freedoms. We also continue if we need the data to establish, exercise or defend legal claims. We do not use personal data for direct marketing.

Withdrawal of consent (Art. 7(3))

Where we process data on the basis of your consent, you can withdraw it at any time. Withdrawing does not affect the lawfulness of what we did before (Art. 7(3) GDPR). For Eclose Omics, your recovery email rests on your consent: “Remove” on your account page withdraws it at once.

Complaint to a supervisory authority (Art. 77)

You have the right to complain to a supervisory authority (Art. 77 GDPR), in particular in the EU country where you live, where you work, or where you think the infringement took place.

How to use your rights

Write to the contact address in section 1. If you have an Eclose account, you can also use the “Your data” section of your account page, which records a request for a copy of your data or for erasure of your account. We answer it within one month. You do not need to give a reason. We may ask for information to confirm who you are (Art. 12(6)). We act only on requests that come from you.

We answer within one month of receiving your request (Art. 12(3) GDPR). If the complexity or the number of requests makes that hard, we may extend it by two more months. We then tell you within the first month, with the reasons. Answering is free, unless a request is manifestly unfounded or excessive (Art. 12(5)).

10. Do you have to give us your data?

Eclose Omics

To use Eclose Omics you need an account. Your user id, your access token, and the files and settings of the jobs you run are needed to enter into and carry out our agreement with you. The security log and rate limiting are part of running the service. Without them we cannot give you access. No law requires you to provide them.

Feedback through the feedback tool is optional (it is sent only if you agree), and not sending any has no effect on the service. Emailing the contact address is optional too, but without an email we cannot answer you. The recovery email is optional as well: without it, if you lose your token, we replace it for you instead of you doing it yourself.

PubVia

You do not have to give us personal data to use PubVia. You can turn off automatic online search and error reports in Settings. The AI features send data only after you set up a provider. The update check carries nothing about you.

Orasis

You do not have to give us personal data to use Orasis. Problem reports are off unless you turn them on, and you can turn off the automatic update check. To download Orasis you need the personal link we send you, so we record your downloads.

Eclose AI (the chat and Eclose Folio)

You do not have to use the chat or Eclose Folio. If you use them, the AI needs the data in section 3 to answer. The chat reads your Omics analyses only if you turn that switch on. In Eclose Folio you can turn off “Show full file paths to AI”.

11. Automated decisions

We make no decision about you based solely on automated processing, including profiling, that has legal effects on you or similarly significant effects (Art. 22 GDPR). Eclose Omics does run automatic technical checks. They test whether an input is valid, and the quota, spending and rate limits we set for each account. For bulk RNA-seq, they also test whether a library looks like full-length or 3′-end RNA-seq. These checks concern the data and requests you submit. They are not an assessment of you as a person.

PubVia

PubVia makes no decision about you either. Its checks — including the Checks judge, which your AI provider runs on a sentence and up to six passages — assess the citations in your text, not you.

Orasis

Orasis makes no decision about you either. It analyses your images, not you.

Eclose AI (the chat and Eclose Folio)

Eclose AI makes no decision about you either. The AI answers your requests. It does not assess you.

12. Browser storage and cookies

Eclose Omics

The /results page keeps your token in the session storage of your browser tab, under the key aspis-console-token, so you stay signed in while the tab is open. It is removed when you sign out or close the tab, and it is sent only to our server, in the requests the page makes.

The account area sets one cookie: the session cookie of the sign-in in section 3. Its name is __Host-eclose_account. It holds an opaque identifier, not your token. It is marked HttpOnly (no script can read it), Secure (sent only over https), SameSite=Lax (not sent with requests from other sites), and Path=/. Other sites cannot set a cookie with that name for us. It is sent only to our server. It is cleared when you sign out or the session ends.

We use no analytics, advertising or tracking tools. We load no fonts, scripts, images or other files from third parties. One exception: the access request form and the sign-in page load Cloudflare's Turnstile script. It runs the check described in section 13. Every other page loads files only from our own server. The storage described here is strictly necessary for the page you asked for. So we show no consent banner.

PubVia

PubVia keeps its data on your computer, not with us. The API keys you enter, and your ChatGPT sign-in, are kept in the macOS Keychain, under “com.pubvia.app.ai-keys”. Settings are kept in a plain-text file, ~/.pi/agent/.paper-lab-keys.json. Besides your choices, the file holds the name you use for comments and the path of the last project folder you opened. Versions before 0.1.50 could leave API keys in that file in plain text. Version 0.1.50 removes them when it starts. The folder ~/Library/Application Support/PubVia holds the app's log, the cache of your library and of online search results, the record of past checks, Word recovery copies of your documents, your chat history and your memory notes. macOS keeps the data of the app's window in ~/Library/Caches/com.pubvia.app and ~/Library/WebKit/com.pubvia.app. Each project folder you open gets a .pubvia folder with the citation cache and attachments. Only the Keychain is encrypted.

PubVia reads your Zotero library only on your computer, through Zotero's local connector and a temporary copy of its database. The app contains no analytics or tracking tools.

The app has no command that deletes all of this. To remove it, quit PubVia. Delete the three folders in ~/Library named above, the settings file, and the .pubvia folder in each project. In Keychain Access, delete the items you find by searching for “com.pubvia.app.ai-keys”.

Orasis

Orasis keeps its data on your computer, in three folders. The first, %APPDATA%\Orasis, holds your settings and privacy choices, and the access token of its local MCP server in a file only your Windows account can read. It also holds the problem reports not yet sent, its server log (up to 2 MiB, plus one older file, with paths and data file names replaced), and the record of what you shared with your AI client, with file paths (that record is never sent). Your list of export folders and the list of runs you chose to keep are there too. The second, %APPDATA%\orasis-standalone, is the app's own data folder. It holds the runs and previews, the saved macros, the training sets, the regions you drew, the lab book, the files downloaded by link and the viewer caches. It also holds the crash dumps (the newest 10, never uploaded) and an optional settings file you can add for Fiji and folder settings. The third, %LOCALAPPDATA%\Orasis, holds the AI code's scratch folders and the sandbox's grant records. Nothing is encrypted at rest. Report bundles and exports go to the export folder you choose, and an export can include your Windows user name.

Orasis sets no cookies. It contains no analytics, telemetry or crash-upload tool. Bio-Formats' own upgrade check is switched off in Orasis, so Orasis never contacts the Bio-Formats project (OME).

To remove Orasis's data, uninstall Orasis, then delete the folders %APPDATA%\Orasis, %APPDATA%\orasis-standalone and %LOCALAPPDATA%\Orasis, and your export folder.

Eclose AI (the chat and Eclose Folio)

Eclose Folio keeps its data on your computer, not with us: its settings, the local key of its MCP server, the page key file with the port and key of a running session, the timestamp of your last session, a copy of the app guide, your lab-book files (.md files in your lab folder) and the undo journals in your library folder. It writes no log files. It has no telemetry, no crash reports and no update checks. It keeps only the refresh token of your sign-in, in your operating system's credential store.

Removing the app leaves its settings folder (~/Library/Application Support/eclose-organizer on macOS, %APPDATA%\eclose-organizer on Windows) and your own files. You delete them by hand.

13. How we protect your data

14. Changes to this notice

We may update this notice, in particular when a processor, a retention period or a category of data changes. The version in force is the one published at this address, identified by the date above. If we plan to use your data for a purpose not stated here, we will tell you before we do (Art. 13(3) GDPR).

Last updated: 2026-11-18.

Terms of Use · eclose.bio