Home Product Solutions Pricing Measured accuracy Security & architecture Compare with other tools Developers VS Code and JetBrains plugins Contact Sign in

Legal & Privacy

Privacy Policy

This Privacy Policy explains how the operator of filterit ("Operator", "we", "us") collects, uses, discloses, and protects personal data when you use the AI chat and data-masking gateway service (the "Service", "Filterit"). It should be read together with the Terms of Service, the Acceptable Use Policy, the AI Disclosure, the Cookie Notice, and the Sub-processors list.


1. Data Controller and Contact

1.1 The controller of your personal data is the operator of filterit (legal name and registered address available on request at contact@filterit.ai).

1.2 General privacy contact and data-protection contact: contact@filterit.ai. No Data Protection Officer has been appointed.

1.3 You also have the right to lodge a complaint with a supervisory authority: in Greece, the Hellenic Data Protection Authority (HDPA / Αρχή Προστασίας Δεδομένων Προσωπικού Χαρακτήρα), www.dpa.gr, or with the authority in your EU country of residence.


2. Important Note on Pseudonymisation (Not Anonymisation)

2.1 The Service replaces detected identifiers in your content with reversible tokens (e.g. [NAME_1]) before sending content to third-party AI providers. This is pseudonymisation, not anonymisation. Detection runs on our own servers in the EU on the text exactly as you submitted it; only the masked text is sent to the AI provider. If you select the explicit unmasked mode (RAW) for a request, that request is sent to the AI provider without masking, as the warning on screen states.

2.2 A token to original-value mapping is retained, encrypted with AES-256-GCM, in a secure store (the "Vault"). Because the original values can be restored, the content remains personal data under the GDPR throughout. We do not claim that "no personal data leaves your server"; rather, masked/pseudonymised content (tokens in place of detected identifiers) is what is transmitted to third-party AI providers for inference. The mapping is deleted automatically 30 days after it was created (Section 7).

2.3 Detection is best-effort. Automated PII detection can miss or misclassify data (see AI Disclosure and ToS Section 8). You remain responsible for not submitting personal data you are not authorised to share.


3. Categories of Personal Data We Collect

Category Examples Source
Account data Account e-mail, optional display name, hashed password; for Google sign-in the Google account identifier and the profile-picture address; time of registration and of last sign-in You
Chat content Prompts and AI responses. Stored masked and encrypted at rest (AES-256-GCM), un-masked only on read for you You
Uploaded files & OCR Documents/images you upload and text extracted via OCR. The files themselves are never written to the database: they are held on temporary server storage for up to 30 minutes (originals used for visual redaction: 10 minutes) and then deleted; the extracted text is kept in memory only while it is processed. Only the masked extracted text is stored, encrypted You
Vault data Encrypted token to value mappings (AES-256-GCM), deleted automatically after 30 days, plus HMAC-indexed search terms (a database administrator cannot read the values) Derived from your content
Email Assistant data (opt-in only) OAuth access and refresh tokens (encrypted) and the address of the connected mailbox. Threads you open in the assistant are fetched live and not stored. If you switch on the mail automation for a mailbox, each processed message is stored masked and encrypted (subject, sender, summary, masked thread, masked draft) until you disconnect the mailbox or delete your account You / Google or Microsoft
Subscription data Plan, billing status, Stripe customer reference. Payments are processed by Stripe on Stripe-hosted pages; card details never reach us You / Stripe
Technical/log data Web-server access logs (IP address, time, path, status; the session cookie, authorisation headers and one-time tokens are removed before writing), application logs, and per-IP rate-limit counters held in memory for up to 24 hours Automatic
Cookie data One strictly necessary session cookie, auth_session (HttpOnly, SameSite=Lax, Secure), and browser local storage for your preferences (theme, language, masking mode). Cookieless usage measurement (Section 2.5 of the Cookie Notice): page path without identifiers, referring site, browser, operating system, device type, screen size, language, country and approximate city from the IP address, which is not stored; the visitor is a daily hash. No third-party analytics provider Automatic
Egress record A PII-free entry for every send to an AI provider (destination, time, size, SHA-256 of the masked request, masked preview encrypted) Automatic
Consent records Document version, time, IP address and browser identification when you accept the Terms, this Policy or the DPA Automatic
Agent download register IP address and browser string for each download of the desktop agent, blanked after 90 days Automatic
Browser extension (only if you install it) The text of a prompt to ChatGPT, Claude, Gemini or Copilot is sent over TLS to the zero-retention masking endpoint (or to your own instance, if you set one) and comes back with placeholders; we do not store it. The placeholder to value mapping stays in your browser session and is cleared when the browser closes. Your API key is kept in the extension's storage in the browser profile. In organisations the extension may be installed by the device policy (Chrome, Edge); the browser then periodically asks our server for the extension's update manifest, without user data You
Web search queries Queries the AI model writes when web search is on in a conversation; placeholders are removed from them and the query is then run through the same detection as a message, so a detected value is removed before it leaves; sent to public search engines through an aggregator we run; recorded in the egress log; not stored by us Derived from your conversation

3.1a Demo without an account. The home page offers a short masking demo (up to 1,500 characters). The text is processed on the spot on our EU servers, shown only on your screen and not stored (neither the text nor the mappings). We keep only per-IP call counters for up to 24 hours to limit abuse.

3.2 We design for zero raw-PII-in-database: chat messages, uploaded text, and email drafts are stored in masked form; original values exist only inside the encrypted Vault, for at most 30 days.

3.3 Customer-held key. For individual accounts that sign in with a password, the key protecting your Vault and content is itself wrapped under a key derived from your password. It is unlocked when you sign in and stays in server memory until 120 minutes of inactivity (at most 24 hours) or a restart; signing out does not remove it. The key is not customer-held for organisation workspaces or for accounts that sign in only with Google: those are protected under our operator key, and when your key is absent new Vault entries are also written under the operator key. While a key is unlocked, or where the operator key applies, we are technically able to read the data; we do so only to operate the Service.


4. Purposes and Legal Bases (GDPR Article 6)

Purpose Legal basis
Creating and administering your account; providing the chat/masking pipeline; storing your sessions Contract (Art. 6(1)(b))
Operating the Vault to mask and later un-mask your content for you Contract (Art. 6(1)(b))
Security, fraud prevention, abuse detection, fail-closed safety, service integrity and improvement Legitimate interests (Art. 6(1)(f))
Connecting your Gmail/Outlook mailbox via the Email Assistant Consent (Art. 6(1)(a)), granted via OAuth, withdrawable at any time
Processing payments and managing subscriptions Contract / legal obligation (e.g. tax records)
Complying with legal obligations and responding to lawful requests Legal obligation (Art. 6(1)(c))
Measuring the use of the site and the app in aggregate, without cookies and without a third-party provider (Section 2.5 of the Cookie Notice) Legitimate interests (Art. 6(1)(f)): knowing which pages and features are used; the browser's Global Privacy Control or Do Not Track signal is honoured

4.2 Where you input special-category data (Art. 9 GDPR), you are responsible for ensuring an Art. 9 condition applies (e.g. your explicit consent or that of the data subject), as stated in the Terms.

4.3 You may withdraw consent (e.g. disconnect the Email Assistant) at any time without affecting prior lawful processing.


5. Recipients and Sub-processors

5.1 We share personal data only with the following categories of recipients, under appropriate contractual and technical safeguards:

  • Third-party AI provider (LLM inference on masked content): Anthropic, PBC (Claude API), United States. It receives masked/pseudonymised content only.
  • Hosting / infrastructure: Hetzner Online GmbH, Germany (data centre Falkenstein). Database backups are kept on the server for up to 14 days.
  • Email providers (opt-in only): Google (Gmail) and Microsoft (Outlook/Graph), only for Users who connect a mailbox via the Email Assistant. Google LLC also acts as identity provider when you choose Sign in with Google.
  • Database/cache infrastructure: a database and a cache operated by us. The cache keeps, for up to 30 minutes, the encrypted form of Vault rows for fast un-masking (decrypted only in the application's memory); the database holds the encrypted Vault and the encrypted content.
  • Payment provider: Stripe, where you purchase a paid plan (account e-mail, plan and an internal reference; card data is handled by Stripe only). Transactional e-mail: Brevo (SMTP relay) for verification codes, password resets, invitations and contact-form messages; it receives the recipient address and the message.
  • Public search engines (via a search aggregator we run): model-written search queries, only when web search is on in a conversation; it is on by default and can be switched off in the chat settings. When the model decides to search, it writes the query itself; placeholders are removed from it and the query is then run through the same detection as a message, so a value the detection recognises is removed before the query leaves. The query goes from our server through the aggregator to public search engines (Google, Bing, DuckDuckGo and others in the default set), the results come back to the model, and each query and each page fetched is written to the egress record (destination, size, hash, masked preview). Detection is not infallible: a value it does not recognise can still be part of a query.

5.2 The current, detailed list is maintained on the sub-processor page.

5.3 Reservation clause. We may engage additional or replacement sub-processors. We will update the sub-processors list and, where required, provide notice and an opportunity to object. Anthropic is the only AI provider we use at present; we do not use OpenAI, Google Gemini or GitHub Copilot.

5.4 We may also disclose data to comply with law, enforce our Terms, or protect rights, property, or safety.


6. International Transfers

6.1 Hosting (Hetzner, Germany) and transactional e-mail (Brevo, France) take place in the EU and involve no third-country transfer. Anthropic is established in the United States, Stripe may process payment data there, and Google and Microsoft (opt-in mailbox connection, Google sign-in) may also process data outside the EEA. For those transfers we rely on the Standard Contractual Clauses contained in each provider's data processing agreement, with pseudonymisation of the content as a supplementary measure where content is involved.

6.2 You may request information about these safeguards at contact@filterit.ai.


7. Retention

Data Retention
Account data For the life of your account. When you delete the account in Settings, the profile, conversations, Vault entries, projects, mailbox connections and consent records are deleted at once
Chat content / uploaded text (masked) Until you delete the relevant session/message or close your account (stored masked and encrypted)
Vault entries (encrypted) 30 days from creation, then deleted automatically (daily job). After that, placeholders in older conversations can no longer be restored to their values. Deleting a conversation does not delete Vault entries, because they are shared across your conversations; deleting your account deletes them at once
Email Assistant Threads: not stored (fetched live). Drafts: masked and encrypted, deleted automatically 90 days after creation or earlier when you delete them. Mail-automation items (opt-in): masked and encrypted, kept until you disconnect the mailbox or delete your account
Email OAuth tokens Until you disconnect the mailbox or close your account
Server logs (may include IP) Web-server access logs: 7 days. Application logs: rotated by size and not kept longer than 30 days. Rate-limit counters: up to 24 hours
Session cookie 8 hours from your last request (renewed while you are active); removed at sign-out
Consent records For the life of the account, deleted with it
Agent download register IP address and browser string blanked after 90 days; the row stays without them
Egress record PII-free, retained as an integrity record
Usage measurement Aggregate visit figures, without an IP address and without an identifier that could link to a person, kept for statistical purposes
Demo text Not stored
Backups Database backups are kept on the server for up to 14 days

7.2 We may retain limited data longer where required to comply with legal obligations, resolve disputes, or enforce agreements.


8. Your Rights (GDPR)

8.1 Subject to conditions and exemptions in the GDPR, you have the right to: access; rectification; erasure ("right to be forgotten"); restriction of processing; data portability; object to processing based on legitimate interests; and withdraw consent at any time.

8.2 How to exercise. Many rights are available self-service in Settings (Export my data, Delete account). You may also contact contact@filterit.ai. We will respond within the statutory time limit (generally one month).

8.3 You will not be subject to a decision based solely on automated processing that produces legal or similarly significant effects; the AI Output is informational and you remain responsible for any decisions you make.


9. Security Measures

9.1 We implement technical and organisational measures appropriate to the risk, including:

  • AES-256-GCM encryption of Vault entries (token to value mappings) and of Email Assistant OAuth tokens;
  • AES-256-GCM encryption at rest of stored content (messages, titles, project texts, drafts, mailbox credentials); a customer-held key for individual password accounts (Section 3.3); a PII-free, hash-chained record of every send to an AI provider;
  • HMAC-indexed search, so search functions operate without exposing plaintext to the database (a database administrator cannot read original values);
  • a fail-closed masking pipeline: if a PII detector is unreachable, the request is refused rather than sent un-masked;
  • a zero raw-PII-in-database design (stored content is masked);
  • access controls, hashed passwords, encrypted transport, and short-lived in-memory caches.

9.2 No system is perfectly secure; we cannot guarantee absolute security.


10. Personal Data Breaches

10.1 We maintain procedures to detect, investigate, and respond to personal-data breaches. Where required, we will notify the competent supervisory authority without undue delay and, where feasible, within 72 hours of becoming aware (GDPR Art. 33), and affected individuals where the breach is likely to result in a high risk to their rights and freedoms (Art. 34).


11. Children

11.1 The Service is for adults: you must be 18 or older. We do not knowingly collect personal data from anyone under 18. If we learn that we have, we will delete it.


12. United States: California Privacy Notice (CCPA/CPRA)

12.1 This Section applies to California residents under the California Consumer Privacy Act, as amended by the CPRA.

12.2 Categories of personal information collected in the past 12 months: identifiers (email, optional display name, IP address, Google account identifier and profile-picture address); account credentials (hashed password); internet/electronic activity (server logs); commercial information (subscription/plan); and the content you submit (which may contain personal information about you or others), stored in masked form. Sources and purposes are described in Sections 3 to 4.

12.3 We do not sell or share your personal information as those terms are defined under the CCPA/CPRA, and we have not done so in the preceding 12 months. We do not use or disclose sensitive personal information for purposes requiring a right to limit.

12.4 Your California rights: to know/access, to delete, to correct, to opt out of sale/sharing (not applicable, as we do not sell or share), to limit use of sensitive personal information, and to non-discrimination for exercising your rights.

12.5 How to exercise: via account settings or contact@filterit.ai. We will verify your request as required by law; you may use an authorised agent.


13. Changes to This Policy

13.1 We may update this Policy. For material changes we will provide reasonable notice (e.g. email or in-app). The "Last updated" date reflects the latest version.


14. Contact

Privacy questions: contact@filterit.ai. Data-protection contact: contact@filterit.ai. No Data Protection Officer has been appointed. Controller: the operator of filterit (legal name and registered address available on request).

Pseudonymisation of personal data in Greek and English.

Product

  • Product
  • How it works
  • Pricing
  • Measured accuracy
  • Security & architecture
  • Compare with other tools
  • Contact

Solutions by sector

  • Law firms
  • Accounting firms
  • HR teams
  • Clinics & practices
  • Public sector
  • Guide: GDPR checklist for AI (Greek)
  • Guide: AI without leaks for lawyers (Greek)

Developers

  • Masking API
  • Python and Node SDKs
  • MCP server
  • VS Code and JetBrains plugins
  • Browser extension
  • Word add-in
  • Desktop agent

Legal

  • Terms of Service
  • Privacy Policy
  • Acceptable use policy
  • AI Disclosure
  • Cookies
  • Sub-processors
  • Data processing agreement (DPA)

filterit detects and pseudonymises or redacts the personal data it recognises. No automated tool can guarantee complete coverage: our accuracy is measured and published, and the tool improves with every release. Review the findings before you send a text, submit data on your own responsibility and only where you are entitled to, and use the service for lawful purposes only. Terms of Service, Acceptable use policy.

© 2026 filterit. All rights reserved.

ChatGPT, Claude and other product names belong to their respective owners. filterit is not affiliated with or endorsed by them.