Security & data handling

Built to be trusted with your accounts.

Ward can read your email, files, and calendar and act on your behalf — so how it handles access and data matters. Here's exactly how it works today. Ward is in early access; wherever something isn't in place yet, we say so plainly.

Connections

Your accounts connect over OAuth — and stay revocable.

OAuth, through the provider

You connect an app through its own consent screen (Google, Notion, and others). You approve the access; Ward receives a scoped token — never your password.

Tokens encrypted at rest

Connection tokens and secrets are encrypted before they're written to the database, so a leaked backup or snapshot yields ciphertext, not live credentials.

Disconnect any time

Revoke a connection from Settings whenever you like and Ward stops using it. The assistants themselves never see the raw credentials.

Grant only what's needed

Keep an assistant read-only, or require your approval before it changes anything — your call, per assistant.

Isolation & access

Your workspace is walled off at the database.

Row-level isolation

Each workspace's data is separated by Postgres row-level security — enforced inside the database, so no query can read across workspaces, even by mistake.

Verified sign-in

Authentication runs on Auth0. Every API request is checked against a verified token before it can touch your data.

The AI models

We don't train on your data.

Ward runs on commercial large language models hosted on Microsoft Azure. We never train AI on your data — and under Azure's commercial API terms, data sent through the API isn't used to train the models either.

What assistants can do

Nothing consequential happens without a record — or your OK.

Approve before acting

Anything that sends, posts, or changes something can be gated behind your approval. Unattended runs skip un-approved actions rather than guessing.

Every action is logged

Each durable step an assistant takes — what it wrote, where, and whether it worked — is captured in an audit trail you can inspect, run by run.

Safe by default on the open web

When an assistant fetches a web page, the request is filtered to block internal systems and cloud metadata and forwards no credentials — so a malicious page can't pivot inward.

Look, don't touch

Run any assistant read-only: it can read and report, but never act.

Hosting & data

Where your data lives, and how to remove it.

Runs on Microsoft Azure

Ward's application and database run on Microsoft Azure.

Yours to remove

Delete an assistant and its records from the app any time, or disconnect an integration to revoke access. Want your whole account and its data removed? Email us and we'll take care of it.

Being straight with you

What's still in progress.

Ward is in early access. Our privacy policy is published and covers what we collect, what we log, and what we never do with your data. A Data Processing Agreement for teams and third-party audits like SOC 2 aren't in place yet — they're on the roadmap as we move toward general availability. If you're evaluating Ward for an organization and need specifics or paperwork, get in touch and we'll tell you exactly where things stand.

Questions about security?

We'd rather answer them than have you guess.