Privacy policy
Effective October 7, 2026 · Operated by Ryan Rudd
WarnBird is an early warning service for suspicious requests and patterns in your connected email. This policy explains how we handle information when you visit our website, create an account, connect an inbox or use alerts.
Information we access
We receive account details through Clerk, our authentication provider, including your identifier, name and email address. With your permission, we access connected email sender and recipient information, subjects, message text, timestamps, thread context and message metadata. We also receive OAuth credentials that allow monitoring while your inbox remains connected.
Gmail access uses the read-only Gmail permission. It does not allow WarnBird to send, change or delete your Gmail messages. The initial scan covers the previous seven days. Ongoing checks are bounded by processing limits and service availability. Attachments, images and linked websites are not comprehensively inspected.
If you use manual checks, we process the information you submit. Optional alerts require contact details and your preferences. Billing is handled by Stripe; we receive subscription and transaction identifiers and status, rather than your complete card number. Our hosting services process request metadata, such as IP addresses and operational logs, to deliver and secure the service.
How we use information
We use connected email information to provide your monitoring, identify warning signs and connections between messages, display supporting evidence, and deliver alerts you enable. We use account and operational information for authentication, billing, reliability, security and support.
We do not sell Google user data, use it for advertising, or use connected email content or its derivatives to train a general-purpose AI model. Monitoring permission does not authorize reuse of your messages in a shared training or evaluation collection. Human access to Google user data is limited to your specific consent for support, necessary security investigations, legal obligations, or internally aggregated operations consistent with Google's Limited Use requirements.
AI processing and service providers
Our monitoring servers process email data. Selected, bounded email evidence is sent to OpenRouter and TypeSafe AI's Jev service for classification. This can include sender and reply-to information, a subject, selected message text, link representations and authentication observations. Recognized sensitive formats are redacted where possible, but redaction does not remove every personal detail and is not anonymization.
OpenRouter and TypeSafe publish commitments not to train models on submitted inputs. These commitments do not mean zero retention. Their processing and retention practices are described in the OpenRouter privacy policy and TypeSafe privacy policy. We do not claim that every provider deletes inputs immediately.
We use Clerk for authentication and encrypted inbox connection metadata, Vercel for the website, Hetzner for monitoring infrastructure, and Stripe for billing. Where configured and enabled, Amazon SES delivers email alerts and Twilio processes phone verification, text and call alerts. Alert services receive recipient destinations and bounded warning content necessary for delivery. A trusted contact receives eligible warning information only when the relevant sharing is enabled and consented to.
Service providers process information to provide these functions. Information may be processed in the United States and other countries where these providers operate. We may disclose information when legally required or necessary to address a security incident.
Storage, security and retention
OAuth credentials and designated private evidence fields are encrypted in application storage. We use HTTPS for communications. No system can guarantee absolute security.
The normal monitoring path does not persist full message bodies, raw HTML, attachments or model prompts in our monitoring database. It retains derived findings, classifications, provider identifiers, coverage and limited encrypted dashboard context, such as a redacted subject and claimed sender details.
While healthy cleanup workers run, observations, analyses and findings are generally removed after approximately 30 days. Decision caches expire approximately 30 days after their last write. Manual checks generally expire after seven days. Usage and cost metadata, account preferences, billing records and delivery records have separate operational retention. These periods are not an instantaneous deletion guarantee. Infrastructure backups and third-party records may persist after active application records are removed.
Your choices and deletion requests
You can disconnect an inbox in Connections to stop future acquisition. You can also revoke Google's authorization at Google Account connections. Disconnecting is distinct from closing your account, cancelling billing or erasing every stored record.
To request access, correction, deletion or account closure, email ryanruddd@gmail.com from your account email. Please do not send passwords, authorization tokens or full email contents. We verify ownership before processing a request, explain records that must be retained for legal, billing or security purposes, and coordinate applicable provider requests. Deletion is handled through a support process; it is not an instant self-service operation. Removing WarnBird data does not delete your original email or recall an alert already delivered.
Google API data
Our handling of information received from Google APIs is governed by the Google API Services User Data Policy, including its Limited Use requirements. Google authorization does not permit unrelated advertising, sale of email data or general-purpose model training.
Changes and contact
We will update this page when our practices change and identify the effective date. For privacy questions or requests, contact Ryan Rudd at ryanruddd@gmail.com.