Built for teams that answer to auditors.

Customer text

Text you check is used to produce findings. Flagged sentences and their verdicts also retrain the classifier that runs inside the browser extension, stored without your name or email. Your organization can opt out.

Model providers

Anthropic and Google process the text of a checked sentence through their enterprise APIs to produce a verdict. Neither trains on it.

Hosting

Google Cloud. The platform and its database run in us-east4, Google model calls in us-central1, and retrieval of regulatory text in europe-west2. Retrieval queries carry the text being checked.

Access control

Sign-in through Google. Role-based access within each organization, and shared access only after the invited person accepts.

Audit trail

Append-only. Every ClaimGuard flag, applied fix and dismissal is recorded with who acted and when. An integrity check runs over the ledger every 15 minutes.

Integrity incidents

A failed integrity check records an alert, exports a forensic snapshot, and opens a draft incident report for the team to complete.

The extension examines your text on your own machine first, with a classifier that runs inside the extension. Only a sentence it marks as a possible regulated claim is sent to Mutaris, with the sentence before and after it for context.

• A sentence is truncated at 1000 characters, and at most 25 are sent in one request.
• The extension does not send the page you are on, its address, your browsing history, keystroke timing, screenshots, or files.
• Checking is off on any site until you turn it on, and you can turn it off everywhere from the popup.
• The Word add-in sends a document’s sentences only when you press Check document. It does not upload the file.

The browser extension and the Word add-in each carry their own short-lived sign-in token, which expires after 12 hours. Neither token carries your organization: it is resolved on every request, so removing a member takes effect immediately.

Every 15 minutes, Mutaris recomputes a SHA-256 hash tree over the audit ledger and compares it with the last known-good baseline. A deleted or altered entry breaks the match. This is an internal integrity check over Mutaris’s own ledger, not a blockchain mechanism, and no outside party can verify it independently.

• Audit records are append-only and are kept for the period the relevant regulation requires.
• Documents and audit results are kept until you delete them or close the account.
• Classifier training records have no automatic expiry yet. Setting one is open work.
• You can ask for a copy of what we hold about you, or ask us to correct or delete it, and we act on the request within 30 days.

Report a security concern or ask a question at hello@mutaris.ai. Privacy requests go to privacy@mutaris.ai.