Your data, your control

Privacy Policy

What we access, how we use it, and the things we'll never do with your data.

Last updated: 3 August 2026

Sylvia ("the Extension") is a Chrome extension that helps you manage your Gmail inbox using AI. This policy explains what data we access, how we use it, and what we never do with it — in plain language, without legalese. It covers both the free plan (where AI calls go directly from your browser to the provider you configure) and the paid Sylvia AI plan (where AI calls are proxied through our server to Cerebras AI).

1 What data we access

To function, Sylvia requests access to your Gmail account via Google's official APIs. This may include:

  • Email metadata — sender, subject, date, and labels
  • Email content, only when required to carry out a task you explicitly request (e.g. summarise, draft a reply)
  • Your Google account email address, for authentication only

2 How we use your data

All data accessed through Gmail is used exclusively to perform the actions you request within the Extension — search, filter, label, summarise, draft replies. Specifically:

  • Email content is only ever sent to an AI provider when you explicitly trigger an AI action (e.g. "draft a reply", "summarise this thread"). Browsing your inbox does not transmit any content.
  • On the free plan, AI requests go directly from your browser to the AI provider you have configured (OpenAI, Google Gemini, or Azure OpenAI). Our servers are not involved in those requests.
  • On the paid Sylvia AI plan, AI requests are routed through our Vercel-hosted server proxy and forwarded to Cerebras AI. This is necessary to protect the API key from being exposed in the browser. The proxy does not log, store, or inspect email content — it only forwards the request and returns the response.
  • No email data is used for advertising, analytics, or any purpose beyond fulfilling your request.
Email content is never stored. Whether on the free or paid plan, no email body or subject is ever written to a database, log, or file by Sylvia.

3 What we never do

  • We do not sell your data to third parties.
  • We do not store your emails or their content on any server.
  • We do not use your data to train AI models.
  • We do not share your email content with anyone except the AI provider processing your request (Cerebras on the paid plan, or your configured provider on the free plan).
  • We do not log your email address in plain text — only an irreversible HMAC-SHA256 hash is used as an internal identifier.

4 Google API Services

Sylvia's use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Google user data is not used to train generalised AI/ML models and is not shared with any third party beyond what is necessary to provide the features you request.

5 Sylvia AI (Paid plan)

When you purchase AI replies on the paid Sylvia plan, the following applies:

  • AI provider: Cerebras AI. All AI requests on the paid plan are processed by Cerebras AI (cerebras.ai). Cerebras processes the message content solely to return a response and does not use it to train models. See Cerebras's privacy policy for full details.
  • Server proxy via Vercel (Frankfurt, EU). AI requests are forwarded through our server hosted on Vercel, deployed in the Frankfurt, Germany region (EU). The proxy does not log or store email content. Vercel may capture standard infrastructure metadata (IP address, HTTP status code, response time) in its platform logs as part of normal hosting operations — this metadata does not include email content or message bodies.
  • Reply-credit balance. Your remaining AI-reply balance is stored in our Redis database as a number associated with an irreversible one-way hash of your email address. The actual email address is never stored in the database. The Redis instance is provided by Upstash and is hosted in the Belgium, EU region.
  • Session tokens. After you sign in, we issue a short-lived session token (valid for 1 hour) using an HMAC signature. This token is stateless — it is not stored in our database — and contains only your hashed identity and an expiry timestamp.
  • Authentication. Your Google OAuth access token is verified directly against Google's token-info API to confirm your identity; it is never written to our database or logs.
  • Payments. Payment is handled entirely by Stripe. We do not receive or store your card details. Stripe webhooks send us only a pseudonymous user hash and the number of replies purchased — never your name, card number, or billing address.
  • Anonymous usage analytics. We collect anonymised event counts (e.g. "extension installed", "reply generated") and daily/monthly active user counts. Events are associated only with a client-side hash of your email — never with your email address itself — and are used solely to understand aggregate product usage.
Your email address is never stored in plain text. The only identifier we persist in our database is a keyed HMAC-SHA256 hash of your email, which cannot be reversed without the server's secret key.

6 Third-party AI providers (Free plan)

On the free plan, when you configure an AI provider (OpenAI, Google Gemini, or Azure OpenAI), your requests go directly from your browser to that provider and are processed according to their privacy policy. Sylvia does not control how those providers handle data once it is transmitted.

7 Data protection & security

Protecting your Gmail data is fundamental to how Sylvia is built. We apply the following safeguards:

  • Encryption in transit. All communication with Google's APIs, with our server, and with AI providers takes place over HTTPS/TLS. Email content and credentials are never transmitted in plain text.
  • No server-side storage of email content. Email bodies and subjects are never written to any Sylvia-operated database, log file, or storage system — on either the free or paid plan.
  • One-way email hashing. Your email address is hashed with HMAC-SHA256 using a server-side secret before being stored in Redis. The hash cannot be reversed to recover your email address, even with access to the database.
  • OAuth tokens are not persisted by Sylvia. On the free plan, sign-in is handled by Chrome's built-in chrome.identity API; tokens are managed by Chrome and never written to any Sylvia server. On the paid plan, your Google OAuth token is used only to verify your identity for one request, then discarded — it is not stored in our database or logs.
  • Stateless session tokens. After authentication on the paid plan, we issue an HMAC-signed session token valid for 1 hour. It is not stored server-side; validity is verified cryptographically on each request.
  • Sandboxed local storage (free plan). Your AI-provider API keys and preferences are stored using Chrome's chrome.storage.local API, sandboxed to the Sylvia extension and isolated from web pages and other extensions.
  • Minimum OAuth scopes. Sylvia requests only the Gmail, Contacts, and profile OAuth scopes that its features actively use. No scopes are requested speculatively.
  • Rate limiting. Our server enforces per-IP rate limits on all endpoints to prevent abuse.
  • Limited Use compliance. Sylvia's use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Google user data is not used to train generalised AI/ML models.

8 Data retention

The following summarises what we retain and for how long:

  • Email content: never stored by Sylvia — not in our database, not in logs.
  • Reply-credit balance: stored in Redis as a number keyed to your HMAC email hash, for as long as you have an account or credits remaining.
  • Usage events: anonymised event counts are stored indefinitely; the event log (containing only event name, timestamp, and hashed user ID) is capped at the most recent 1,000 entries.
  • Session tokens: stateless — not stored server-side. They expire after 1 hour.
  • Free-plan local data: API keys, preferences, and local chat history are stored only in your browser's extension storage and retained until you clear them or uninstall the Extension.
  • AI provider data: data forwarded to Cerebras (paid plan) or your configured provider (free plan) is retained according to that provider's own policy.

9 Your controls & revocation

You remain in control of Sylvia's access to your data at all times:

  • Revoke Gmail access. Visit myaccount.google.com/permissions at any time to revoke Sylvia's access to your Google account.
  • Delete locally stored data. Uninstalling the Sylvia extension from Chrome removes all locally stored API keys, preferences, and chat history.
  • Switch or remove AI providers (free plan). You can change or delete the AI provider and API key you've configured at any time from the Extension's settings.

10 Children's privacy

Sylvia is not directed at children under 13 and does not knowingly collect data from them.

11 Changes to this policy

We may update this policy from time to time. The "last updated" date at the top of this page will always reflect the most recent revision. Continued use of the Extension after changes are posted constitutes acceptance of the revised policy.