How Scrubbr protects your data

Scrubbr connects to your email, chat and calendar with your permission. Where a control is still being rolled out, we say so.

7 days free·Cancel anytime

Works with the apps you already use

Scrubbr connects to Gmail, Google Calendar, Google Chat, Slack and, if you choose, Outlook and Teams, with your permission.

  • Gmail
  • Google Calendar
  • Google Chat
  • Slack
  • Outlook
  • Microsoft Teams

The short version

You approve what gets sent

Nothing is sent to another person without your explicit action: a click, or a voice command you confirm.

Some actions that affect only your own account happen automatically or on a single voice request. Scrubbr applies its own labels in Gmail. The voice assistant can mark messages read and create, move or cancel events on your own calendar when you ask.

No AI training on your content

We do not use your content to train AI models. Every model request is routed only to providers whose data policies prohibit training on it.

Attachments are never stored

Scrubbr records only an attachment's filename, type and size. Attachment content is never sent to an AI model.

Security assessments

  • TAC Security, ADA CASA assessor CASA Tier 2 verified
  • AICPA SOC 2 Type I seal SOC 2 Type I · LOE upon request

Effective 4 October 2026 · version 1.6

The full detail is on this page

This page explains where your data lives, who can see it and how we protect it.

The details

Where your data lives, and how it is encrypted

Stored in the United States

Application data and customer data are stored in Supabase (managed Postgres on AWS) in a US region (us-west-2).

Application servers run on AWS in the same region, behind Cloudflare.

Backups are encrypted and managed by Supabase.

AI processing is performed in the United States by Anthropic, OpenAI, and Google. If you are located outside the United States, your data is transferred to and stored in the United States under appropriate safeguards.

Encrypted in transit and at rest

Data is encrypted in transit (TLS 1.2 or higher) and at rest (AES-256 with provider-managed keys).

OAuth tokens carry a second, application-layer AES-256-GCM encryption layer. Its key is held outside the database.

Limited access

Production access is limited to our two founders and granted at the smallest scope needed.

We have not built any internal tool for reading customer messages.

How access works

What Scrubbr reads, source by source

Gmail and Outlook

Message metadata (sender, recipient, subject, timestamps, labels) and message body content. Processed for prioritization, summarization, and drafting.

Slack

Message metadata and message body content from channels and direct messages your account has access to. Processed for prioritization, summarization, and drafting.

Scrubbr uses your personal Slack user token, not a bot token.

Messages you send through Scrubbr appear in Slack under your name, indistinguishable from a message you typed in Slack itself. There is no scrubbr-bot installed in your channels.

Google Calendar and Outlook calendar

Event metadata (titles, attendees, times, locations, descriptions). Processed for triage context and meeting prep. Future events are kept; past events follow the retention window.

Google Chat and Teams

Space, channel and direct-message metadata, and message text from the spaces and chats your account belongs to. Processed for prioritization, summarization, and drafting.

Replies are sent only when you approve them. Scrubbr creates a direct-message space or changes space membership only when you take an explicit action.

Attachments

Never stored.

Scrubbr records only an attachment's filename, type and size. When you open an attachment, we fetch it from the source system and hand it straight to your browser without keeping a copy. Attachments you add when composing stream straight through to Gmail, Outlook or Slack without being persisted.

The one exception: small images embedded inline in an email's body are stored as part of that message.

Card data

Scrubbr never sees your card. Payment is handled directly by Stripe; we receive only a customer ID and subscription status.

Retention and deletion

  • Kept for the life of your account by default. That way search and drafting can use your history.
  • A shorter window is available. You, or your company administrator, can set a shorter window in Settings, from 30 days to 10 years. A nightly job then deletes messages older than the window, together with their search embeddings, extracted commitments and past calendar events.
  • Delete your account at any time. You can delete your account from Settings; deletion is immediate. Verified deletion requests by email, and deletion after a contract ends, are completed within 30 days, except where retention is required by law.
  • Export or delete on request. You can request export or deletion at any time by emailing privacy@scrubbr.ai.
  • Audit log entries are kept up to 365 days. They never contain message content. They include IP address and browser, and serve as security evidence.

Who can access your data

Smallest access needed

Production access is limited to our two founders and granted at the smallest scope needed. Access reviews are targeted quarterly; the most recent completed review was in June 2026.

Your approval comes first

We access the contents of your messages only with your explicit approval, except where we must act on a security incident or a legal obligation.

That is typically when you have asked us to investigate something. For Gmail, Calendar, and Chat data, it means your affirmative agreement to view the specific messages concerned. We do not browse customer data.

For Gmail, Calendar, and Chat data, Google's API Services User Data Policy binds us to the same limit.

No tool for reading your messages

We have not built any internal tool for reading customer messages. Administrative tools return account and usage metadata only, and there is no “view as user” feature.

An audit trail

An append-only audit log records sign-ins, integration changes, messages sent and calendar changes made through Scrubbr, administrative actions and deletions. It is kept for 365 days.

How we build and run Scrubbr

Software development

  • All changes are version-controlled. Continuous integration runs type-checking, tests, dependency vulnerability scanning and secret scanning on every pull request and every change to the main branch.
  • With a two-person team, a second-person review is not currently required before merge.
  • Secrets are held outside source control in AWS Systems Manager Parameter Store (encrypted SecureString parameters), in owner-only files on the production host, and in Supabase's encrypted secrets store.
  • They are rotated on personnel changes and whenever exposure is suspected; a fixed rotation schedule is being put in place.
  • Every deployment has a tagged rollback point, and rollback to the previous build takes under a minute.

Logging, monitoring and backups

  • Application logs and error reports pass through a redaction step that strips message bodies, subjects, tokens and email addresses.
  • Backups are encrypted and retained by our managed database provider.
  • Error monitoring runs continuously; the team is alerted to production errors.

Incident response

If we discover a security incident affecting customer data, we contain it, investigate, and notify affected customers without undue delay and in any event within 72 hours of confirmation.

Report a suspected incident to security@scrubbr.ai.

AI and your content

  • We do not use your content to train AI models. Every model request is routed only to providers whose data policies prohibit training on it.
  • One gateway, a no-training policy. Requests go through OpenRouter, an AI model gateway. Every request carries a no-training, no-data-collection policy.
  • Models in use. Google Gemini models handle classification, summaries, drafting and the assistant. Anthropic Claude models handle drafting and the assistant. An OpenAI embedding model powers search. All three are reached via OpenRouter.
  • Attachments stay out. Attachment content is never sent to an AI model.
  • Voice input. Your browser transcribes it. In Chrome, the browser's speech service (operated by Google) processes the audio under the browser's own terms, and Scrubbr receives only the text.
  • No selling, no ad trackers. We do not sell personal information. We do not use third-party advertising trackers on the marketing site.

Sub-processors

The providers below help us run Scrubbr. Each is engaged under written terms to handle data only for the purposes of running Scrubbr.

  • Amazon Web Services (US)Application hosting, compute, queues.
  • Supabase (US)Managed Postgres, auth, encrypted backups.
  • CloudflareContent delivery network and edge security (DDoS protection, web application firewall, TLS termination). It terminates TLS at its edge and inspects requests to block attacks; content passes through in transit only to deliver and protect it.
  • Sentry (US)Application error monitoring. It receives diagnostic telemetry with personal data redacted before it leaves our systems.
  • OpenRouter (US)AI model gateway. Every request carries a no-training, no-data-collection policy, so it is routed only to providers whose policies prohibit training on it.
  • Google (US)Gemini models for classification, summaries, drafting and the assistant, via OpenRouter.
  • Anthropic (US)Claude models for drafting and the assistant, via OpenRouter.
  • OpenAI (US)Embedding model for search, via OpenRouter.
  • Browser push servicesOperated by your browser's vendor; they deliver notifications you turn on. Payloads are encrypted to your browser, so the push service cannot read them.
  • PostHog (US)Web analytics on the marketing site only. Page views, clicks, and feature-usage events. No message content, attachments, or tokens. Session recording off.
  • Stripe (US)Billing. Card data goes directly to Stripe.
  • Google (sign-in, and Gmail, Calendar, and Chat APIs)Sign-in provider if you sign in with Google; your data source.
  • Slack TechnologiesYour data source.
  • Microsoft (sign-in and Graph API)Sign-in provider if you sign in with Microsoft; Outlook mail and calendar and Teams as a data source if you connect them.
  • Zoom (US), only if you connect itCreates meetings when you schedule them from inside Scrubbr.
  • Granola (US), only if you connect itMeeting-notes source. Notes and summaries only; transcripts and audio never accessed.

We give at least 30 days' notice before a new sub-processor handles customer data, by updating this page and emailing account holders.

Assessments and compliance

Completed

Google CASA Tier 2

Scrubbr has completed Google's CASA Tier 2 security assessment for restricted Gmail scopes, with the accredited assessor's Letter of Validation issued in August 2026.

Aligned

NIST SP 800-171 Rev. 2

Scrubbr's security program is aligned to NIST SP 800-171 Rev. 2.

Audit engaged

SOC 2 Type I

Scrubbr is not yet SOC 2-attested.

A SOC 2 Type I audit is engaged with TAC Security, and the letter of engagement is available on request. When the report is issued, this page will be updated and the report will be available under NDA.

Need this in a contract, or for a vendor review?

If your team needs these commitments in a contract and not on a page, we will sign one. A data processing agreement covering processing scope, sub-processors, security measures, breach notification, and deletion is available on request from privacy@scrubbr.ai.

For questionnaires or vendor reviews from your infosec team, send them to security@scrubbr.ai with your due-diligence document and we'll respond promptly.

Security: security@scrubbr.ai · Privacy: privacy@scrubbr.ai

Legal text lives on its own pages: Privacy notice · Terms

Security questions

Does Scrubbr train AI on my email?

No. Scrubbr does not use your content to train AI models. Every model request is routed only to providers whose data policies prohibit training on it.

Is Scrubbr SOC 2 certified?

Not yet. Scrubbr is not yet SOC 2-attested. A SOC 2 Type I audit is engaged with TAC Security, and the letter of engagement is available on request. Scrubbr has completed Google's CASA Tier 2 security assessment for restricted Gmail scopes, with the accredited assessor's Letter of Validation issued in August 2026, and its security program is aligned to NIST SP 800-171 Rev. 2. When the SOC 2 report is issued, this page will be updated and the report will be available under NDA.

Can Scrubbr send email on its own?

Nothing is sent to another person without your explicit action: a click, or a voice command you confirm. Some actions that affect only your own account happen automatically or on a single voice request: Scrubbr applies its own labels in Gmail, and the voice assistant can mark messages read and create, move or cancel events on your own calendar when you ask.

Can Scrubbr staff read my messages?

We access the contents of your messages only with your explicit approval, except where we must act on a security incident or a legal obligation. We have not built any internal tool for reading customer messages. The database itself can be reached by our two founders, for incident response and approved repairs. That limit is enforced by policy, not yet by encryption, and reads made directly in the database console are not individually logged today.

Does Scrubbr use a Slack bot?

No. Scrubbr uses your personal Slack user token, not a bot token. Messages you send through Scrubbr appear in Slack under your name, indistinguishable from a message you typed in Slack itself. There is no scrubbr-bot installed in your channels.

How do I delete my data?

You can delete your account from Settings at any time; deletion is immediate. Verified deletion requests by email, and deletion after a contract ends, are completed within 30 days, except where retention is required by law. You can also set a shorter retention window in Settings, from 30 days to 10 years. Audit log entries (never message content) are kept for up to 365 days as security evidence.

How do I report a security issue?

Email security@scrubbr.ai. If we discover a security incident affecting customer data, we contain it, investigate, and notify affected customers without undue delay and in any event within 72 hours of confirmation.

Never miss anything that needs you

7 days free·Cancel anytime

Start with: 7 days free · Cancel anytime