Privacy · effective

Sonarwork knows a message was sent. Not what it said.

None

Message contents read

Kept

Activity metadata, so years can be compared

11

Subprocessors, all named

One Slack message, as Sonarwork stores it.

Every field kept from this message

Written down

event_typemessage
occurred_at2026-09-07T21:47:12Z
personU04KJ · Marcus R.
workspaceacme-eng
conversationC07PLT2 · an id, not a name
channel_typepublic channel
threada reply
thread_started_byU07PR · Priya I.
text_length33 characters
mentionedU07PR · Priya I.
message_idSlack's id for it

Dropped at ingest

text"@Priya take a look before standup"
filesdiff.patch, shot.png
channel_name#platform-eng
link_previewsunfurled pages

Dropped where the data arrives, before anything is written. No later stage can read them, and no setting turns them on.

The gate is on the kind of data, not the tool.

Ingest filter · the same rule for every source

Passes the gate4 kinds
occurred_atWhen something happened, to the second
graph.edgeWho talked to whom, by account
review.stateThat a PR opened, was reviewed, merged
meeting.spanThat a meeting ran, and for how long
Dropped at the gate5 kinds
message.textAnything anyone typed, anywhere
code.diffCode, diffs, commit messages
review.bodyPR titles, descriptions, comments
meeting.titleTitles, agendas, invitations, notes
doc.contentDocuments, pages, tickets

No meeting recordings either: the meeting notetaker does not join any meeting yet. When it does, it will be off by default, and anyone can say no.

The clauses.

GDPR Art. 13 & 14 · CCPA

01

Who we are

Sonarwork (“Sonarwork”, “we”, “us”) operates the Sonarwork manager-intelligence platform that helps managers see where their teams need attention. The data controller for customer-workspace data is the customer organization that installed Sonarwork; Sonarwork acts as data processor on the customer’s behalf.

Contact for privacy questions: privacy@sonarwork.com

02

What we collect, and why

Sonarwork collects collaboration metadata from the integrations your workspace connects. The examples below cover the most common ones — new integrations follow the same rule: metadata in, message text never. Specifically:

  • Slack (and similar messaging tools): message timestamps, channel ids (hashed), reaction events, message length, whether a message is in a thread, and who was @-mentioned. Never message text or attachments.
  • Google Calendar (and similar calendar tools): meeting count, duration, attendee emails. Never event titles, descriptions, or attachments by default.
  • GitHub (and similar code platforms): pull-request open / merge / review events, file-count + line-count totals, CI workflow status. Never diff contents, commit messages, or issue body.
  • Meeting notetaker (opt-in only, not live yet): the meeting notetaker does not join any meeting yet, so no meeting audio is collected. When it does, it will join a meeting only when an admin has turned it on and everyone from your workspace in that meeting has agreed. Each person answers once, not before every meeting, and can withdraw at any time. Outside the EU, an admin can choose to count everyone as agreed until anyone says no. The audio will be sent to Deepgram for transcription and the transcript and meeting title to one of our AI providers (Anthropic or OpenAI, see the subprocessors below) for the meeting summary. Off by default.

We also collect account profile data (your name, email, role within the workspace, OAuth tokens for the connected services) and product analytics (which pages are opened and which features are used) to improve the product. Product analytics, and recordings of a sample of sessions with all text hidden, run in your browser only after you press “Accept all” on the cookie banner.

On this public website we also keep daily totals: how many times each page was opened, and how many times its demo and early-access buttons were pressed. These totals use no cookie and store nothing about you, not even your address, so they run for every visitor.

The legal basis under GDPR is legitimate interest (helping managers run their teams effectively) for collaboration metadata, and consent for product analytics. You can withdraw consent for analytics at any time without affecting the core service: .

03

How we use your data

  • Generate signals + briefs: the metadata is transformed by signal evaluators (deterministic code) and an LLM (from Anthropic or OpenAI) into a brief that lands in the manager’s inbox each weekday morning (configurable per workspace).
  • Our AI providers do not train on your data. Anthropic and OpenAI process a request only to return an answer. Their business terms say they do not train on data sent through their APIs unless a customer opts in. OpenAI keeps API request logs for up to 30 days for abuse monitoring.
  • Detect bias: we monitor signal precision per cohort (role, tenure) to ensure no group is systematically mis-served.
  • Improve the product: aggregate (non-identifying) metrics inform our roadmap.

We do not sell your data, do not use it for advertising, and do not share it with third parties beyond the subprocessors listed below.

04

Subprocessors (11)

These are every company Sonarwork sends your data to, what they do with it, and where they hold it.

  1. Claude is the AI model that writes your morning brief, summarizes meetings and answers the questions you ask Sonic. We never send the contents of any message — only the structural facts (who, when, how often). Emails, phone numbers, and risky phrases are stripped before anything reaches the model. Anthropic's business terms say it does not train its models on data sent through its API by default.

  2. Listed because Sonarwork can be set to use an OpenAI model to write the morning brief, summarize meetings and answer the questions you ask Sonic. In production Sonarwork uses Anthropic Claude, and sends nothing to OpenAI unless Sonarwork switches this on. If it is switched on, the same rules apply as above: never message content, and emails, phone numbers and risky phrases are stripped first. OpenAI's published API terms say data sent through the API is not used to train its models unless the customer opts in, and that it keeps API request logs for up to 30 days for abuse monitoring.

  3. The database where Sonarwork keeps everything it holds: activity metadata, signals, briefs, accounts and settings. Data exports wait here for seven days and are then deleted.

  4. Where the Sonarwork website and app actually run. Server logs are kept seven days and never contain message content.

  5. Error reports, so we can fix what breaks. Reports carry no IP address and no email. A recording of a sample of browser sessions is kept only for visitors who press "Accept all" on the cookie banner, with every piece of text and every input hidden.

  6. Runs Sonarwork's scheduled jobs and its rate limits. It starts each job on time and delivers each brief at its set minute. It counts sign-in attempts and other requests, so nobody can send them at speed, and it briefly holds the secret of a two-step sign-in being set up. It never holds a name, an email address, an IP address or any activity: each counter is stored under a one-way hash of what it counts.

  7. Transactional + opt-in email delivery for morning briefs, invitations, and export-ready notifications. Bounces + complaints feed our suppression list.

  8. Billing and subscription management. Your card number goes directly to Stripe — Sonarwork never sees it, stores it, or has access to it.

  9. Product analytics, so we can see which features are used (for example "first brief sent" or "integration connected"). In the browser it runs only after a visitor presses "Accept all" on the cookie banner, records which pages are opened and the events Sonarwork sends on purpose, and never records the text on a page or what anyone typed.

  10. Speech-to-text for the meeting notetaker. Used only when an admin has turned the notetaker on and the people from your workspace in the meeting have said yes, and the notetaker joins no meeting yet. Audio is streamed to Deepgram, transcribed, and discarded after we receive the text.

  11. Object storage for short-lived encrypted notetaker audio. Buckets are private; only Sonarwork can read them.

We notify customers 30 days before any subprocessor is added or changed.

Data sources

Sonarwork reads from the tools your workspace connects — Slack, GitHub, Google Calendar, Outlook and any others you add. Your company agreed terms with them directly, so they are data sources rather than subprocessors under GDPR Art. 28. Sonarwork sends back to them only into your own workspace. It sends three things: a manager’s brief as a Slack message, when that manager chooses Slack delivery; a Slack alert to a person’s manager when one of that person’s signals turns urgent, which is on unless the manager turns off “Tell me right away” in Settings, Notifications; and, on the calendar of the person who acted in Sonarwork, the meetings they add in Sonarwork, and their approved time off and working from home.

05

Data retention

How long each kind of data is kept (a workspace can ask for a shorter window):

  • Activity events (raw metadata): kept for as long as your workspace uses Sonarwork, so you can compare years. Deleted for one person when they ask, and for everyone when the workspace is closed (see below).
  • Signal events: kept for as long as your workspace uses Sonarwork, including cleared ones.
  • Meeting transcripts (when the meeting notetaker is on): 90 days
  • Briefs: indefinite (the manager’s historical record)

Your workspace owner deletes your data in Sonarwork when you ask them to (see Section 7). It is deleted straight away.

When a workspace is closed, at the customer’s request or when its contract or pilot ends, Sonarwork disconnects every tool straight away and deletes everything in the workspace within 30 days. That includes every person who belongs to no other workspace. One record is kept: the tamper-proof log of actions taken in Sonarwork (for example who changed a setting or exported data), which nobody can edit or delete.

06

International transfers

Sonarwork’s primary infrastructure is hosted in the United States. Transfers of personal data out of the EEA, the UK and Switzerland rely on the EU Standard Contractual Clauses, with the UK and Swiss addenda, as set out in section 5 of our Data Processing Agreement. The agreement is available on request.

07

Your rights

Whether under GDPR, CCPA, or similar laws, you have the right to:

  • Access the data we hold about you — at /me you can see what Sonarwork collects about you and the signals about you. Your 1:1 notes and your answers to calendar invitations are not shown there yet: email privacy@sonarwork.com for them.
  • Export your data as a JSON file — request from /me; delivered within 72 hours. Your 1:1 notes and your answers to calendar invitations are not in the export yet: email privacy@sonarwork.com for them.
  • Correct inaccurate data — your profile fields are editable; signal-level disputes go through /me/disputes.
  • Delete your data — ask your workspace owner. Your data page at /me names the owner and their email address. Only the owner can delete a person’s data in Sonarwork. When they do, it is deleted straight away.
  • Pause processing — ask your workspace owner to stop new data collection without deleting anything. The owner can pause everything, or one tool.
  • Object to processing — tell your workspace owner why you object. Your company decides, as the controller of your data, and the owner can pause collection about you.
  • Lodge a complaint with your data protection authority — for EU residents, contact your country’s supervisory authority.

For California residents specifically: see our Do Not Sell or Share page. Sonarwork does not sell personal information.

08

Security

We encrypt data in transit (TLS 1.3) and at rest. Sensitive fields and OAuth tokens are sealed with AES-256-GCM at the application layer. Access is restricted to authenticated employees with role-based permissions; every admin action is logged in an append-only audit trail with chained hashes. An internal red-team operates today. An external penetration test and a SOC 2 Type I audit are both planned: we will name the auditor and the dates here once they are booked, and we will not claim either until it is finished.

09

Children's privacy

Sonarwork is a workplace tool intended for use by adults aged 18 and older. We do not knowingly collect data about anyone under 16. If we learn we have collected such data, we will delete it.

10

Changes to this policy

We will notify customers 30 days before any material change to this policy via email and an in-app banner. The date this version took effect is at the top of this page.

11

Contact

Privacy questions, data-subject requests, or complaints: privacy@sonarwork.com. For EU residents we will respond within 30 days as required by GDPR Art. 12(3).

Every right here is a button in the product.

Know

Every event we hold about you, on your own page.

Export

Everything except your 1:1 notes and calendar replies, as a file you can open, within 72 hours. Email privacy@sonarwork.com for those.

Delete

Ask your workspace owner, named on your page. Once the owner deletes it, it is gone straight away.

Dispute

Wrong signal? Saying so starts a queue a person works.

None of the four depends on which plan your employer bought.