Legal document Ā· Scriptor
Privacy policy
Who we are and how to reach us
Scriptor helps researchers write, revise and submit academic papers. The service is operated by Lior Tsafrir, Tel Aviv, Israel. For questions and requests to access, correct or delete information: lior31197@gmail.com.
Do you have to provide information
There is no legal obligation to give us information; providing it depends on your choice and consent. Without a Google account you cannot sign in, and without uploading materials you cannot receive results. You can stop using the service and ask for deletion at any time.
What we collect and why
Do not upload identifying information about research participants, such as names, ID numbers or identifiable medical data. Upload anonymized data only.
- Name, email address and profile picture from your Google account: identification, sign-in, progress notices and sharing invitations.
- Manuscripts, reviewer comments, journal guidelines, tables and data files you upload, and what you write in the conversation: producing the results you asked for.
- Results and conversation history: saving and continuing your work.
- Details of the paper's authors that you enter or confirm in the authors form (name, email, affiliation and ORCID iD, and when the journal asks, academic degree, the corresponding author's postal address and phone, a present address and CRediT roles): building the title page and the submission. Kept with the conversation and deleted with it. Enter other authors' details only with their consent.
- Comments you write on passages of your paper, and comments from a Word file that you choose to import (including the author name the file shows): kept with the conversation, sent to the AI services with the paper in every run on it (not only when you ask for them to be handled), and deleted with the conversation.
- Degree, academic role and department you give in Settings or in the questions before a submission package is prepared: signing the cover letter in your name. Kept in your profile until you change or delete them.
- The name of the editor-in-chief of the journal you chose, when the journal's public pages list it: proposing to address the cover letter to them by name, only after you confirm it. Scriptor reads only the journal's official pages and sends them nothing about you or your paper.
- Usage data (number of model calls, estimated cost, processing times and errors), without the content of your materials: usage limits, troubleshooting and performance.
- Output quality measures, without the content of your materials: downloads of a result (Word, PDF or ZIP), a thumbs rating and a reason from a fixed list, regenerations, accepted versions, error codes, and one number for how much the manuscript changed in the week after delivery. They are linked to your account, deleted when it is deleted, and used only to improve the quality of the service.
- Your answers to the feedback survey, if you choose to answer it (role, ratings and open answers), with the track, the number of completed runs and the conversations where the survey was shown to you: improving the product. The answers are linked to your account, and the service operator sees them with your name and email. They are not sent to AI services, are included in your data export and are deleted with your account.
- Strictly necessary cookies only (sign-in, access code and language). We use no advertising or tracking cookies.
- Browser notification subscription, if you allowed it: a notice when a result is ready, and when a collaborator in a shared statistical project adds a comment or a change proposal (without its content, at most one notice of each kind per project every half hour).
Signing in to the journal's site for you (automatic status updates)
If you turn on automatic status updates for a submission, Scriptor signs in to the journal's submission site (for now, Editorial Manager only) once a day with the username and password you gave, on your behalf and with your explicit consent. The feature is off until you turn it on in the form on the submission card.
- What is read: only the author dashboard on the site, that is the manuscript number, the status text and the folder name. Scriptor does not press action buttons, and does not change, upload or send anything on the site.
- How the sign-in details are kept: the site address, username and password are stored encrypted (AES-256-GCM) in the database. They are decrypted only in the background check, and written to a file inside an isolated virtual machine that is deleted when the run ends. They are not sent to AI providers, not written to logs and never shown again in the interface.
- What is kept from a check: the status that was read, the time of the check and an outcome code (for example a wrong password or a CAPTCHA). The page content is not kept.
- Deletion: the "Stop automatic updates" button on the submission card deletes the sign-in details at once, and so does deleting your account. After deletion they may stay in the database backup for up to 7 days.
- When the site shows a CAPTCHA, two-step verification or sign-in through another identity provider, Scriptor stops and does not try to get around it. After a wrong password it does not try again until you save a new one, so the account is not locked.
- The terms of some submission sites and publishers forbid third-party access to an account, the use of robots, or sharing a password. Scriptor does not check each site's terms for you. Read the site's terms before turning this on: the decision is yours.
- The site may email you about a sign-in from a new location. We recommend a password you use only for this site.
Use of artificial intelligence
To produce results, the materials you upload and the text you write are sent to third-party AI services, listed in the table below. Results are generated automatically and may contain errors; you must check them before any use or submission. We make no decisions about you based on automated processing.
Anthropic, OpenAI and Google do not train models on content sent through their commercial APIs (for Google, the paid tier of the Gemini API), under their terms. Exception: the Mem0 memory service may use text excerpts it receives to improve its product. Perplexity may also use the search queries it receives to improve its services, so it receives only short queries, never manuscript text.
Sharing a manuscript for editing
A shared paper is visible to the people you invited, including AI-written parts. They also see your name and the names of the other participants.
Only one person edits at a time. For this we keep who is editing now, in which browser tab and until when, without any content. The record is replaced when editing passes to someone else and is deleted with the document.
Participants allowed to answer on a shared conversation can see the comments on its released versions and add their own. Their comments are sent to the AI like yours, and your export includes them.
You can remove a participant's access at any time, and it applies from their next request.
Depositing on Zenodo from Scriptor's account
If you choose automatic deposit of a reproducibility bundle, Scriptor sends Zenodo, a research repository run by CERN, the bundle (analysis code, parameters, results and tables, no input data) and the authors' details: name, affiliation and ORCID iD when known. Email addresses are not sent.
Until you approve publishing, the record is a draft that only Scriptor can see. After you approve, the record and the authors' details are public and kept permanently. Zenodo allows deletion only within 30 days of publishing, and even then a page with the citation remains. So the authors' details cannot be deleted from a published record the way information is deleted in Scriptor.
We keep who approved publishing, when, and which consent text was shown.
Who receives the information
Information goes only to providers that run the service for us, and only to run it. Some store or process data outside Israel, mainly in the USA and the EU. By accepting this policy you agree to this transfer. We will not sell your information or pass it on for marketing.
| Provider | Role | Location | Retention at the provider |
|---|---|---|---|
| Anthropic | Main language model | USA | Up to 30 days, no model training |
| OpenAI | Member of the review council, reviewer in the scientific review debate, and fallback when Anthropic is unavailable | USA | Up to 30 days, no model training |
| Google (Gemini) | Member of the review council and reviewer in the scientific review debate | USA | Per the provider's paid-tier terms, no model training |
| OpenAlex | Open scholarly database: identifying the journal you chose, its metrics and articles it published on a similar topic, and the literature search for writing a manuscript and in the citation search. Only the journal name and short topic search phrases are sent, never the manuscript text | USA | Per the provider's policy |
| Perplexity | Web search (Search API), only when the integration is enabled: checking references not found in the databases, web sources for the literature review, current updates on recommended journals, and fact checking in the review council. Only short search queries are sent: a reference's title and year (without author names), topic keywords, the journal name and short reworded factual claims. No manuscript text, data or account details are sent. Under the Search API terms, the provider may keep the queries and results and use them to improve its services | US company; processing location not stated in the provider's terms | Per the provider's terms; retention period not stated |
| Vercel | Website hosting and private file storage | Files in Frankfurt, Germany; hosting is global | While the account is active |
| Neon | Database | Frankfurt, Germany | While the account is active |
| Inngest | Background job orchestration; step content is encrypted before it is sent | USA | Up to 24 hours |
| Upstash | Request rate limiting: user id or IP address, no content | Cloud | A few minutes |
| Langfuse | Cost and usage measurement, no content | EU or USA | Per the provider's policy |
| Vercel Sandbox, E2B | Running statistical analyses on data files, and an isolated browser that reads the submission status from the journal's site, if you turned on automatic updates | USA; status reading in Frankfurt, Germany | Deleted when the run ends |
| Sentry | Error tracking, no content and no user details beyond the account id | Frankfurt, Germany | Per the provider's policy |
| Email provider (SMTP) | Sending invitations and notifications | Per the provider | Per the provider's policy |
| Mem0 | Writing preference memory: excerpts of your requests are sent to search and save preferences. Under Mem0's free plan terms it may use them to improve its product and train models | USA | While the account at Mem0 is active |
| Zenodo (CERN) | Public research repository, only if you chose automatic deposit and approved publishing: the reproducibility bundle (no input data), the authors' names, affiliations, ORCID iDs, the title and keywords. These details become public | Switzerland | Permanently, as a public record with a DOI |
How long we keep it
- Data files ("Data" role) are deleted automatically 90 days after upload, with reminders 7 days and 1 day before.
- Manuscripts, results and conversations are kept while your account is active or until you delete them.
- The database allows point-in-time restore for the last 7 days, so deleted information may remain in backups for up to 7 days.
Your rights
- Access the information about you (section 13 of the Israeli Privacy Protection Law).
- Ask to correct or delete information that is not correct, complete, clear or up to date (section 14).
- Request export, deletion or a pause in processing from your account settings, or by writing to lior31197@gmail.com.
- Complain to the Israeli Privacy Protection Authority.
- Users in the EU also have GDPR rights, including restriction, objection and portability, and may contact their national supervisory authority.
Security
Access to information is limited to the service operator. Files are kept in private storage and every read is checked against the owner. Traffic is encrypted, and background job content is encrypted before it is sent to the provider. No system is completely secure; if a breach is found, we will notify you and the authority as the law requires.
Changes
We will announce any material change in the service in advance and ask for your consent again before it applies to you.