Security Promise
AI Prompt Leak Guard is built to stop risky prompts before they leave supported AI pages. The check happens in the browser and the user sees the decision before sending.
Browser Permissions
The extension asks only for the access it needs. Required permissions are active tab access for user-invoked scans, local storage for policy settings, content-script registration for supported, explicitly configured hosts, context menus for the selection check, notifications for confirmations, and a Chromium-only side panel. Its 18 built-in AI services use 31 named HTTPS hosts in both content-script matches and optional-host declarations. Browser site access can be granted at install; site switches are protection controls and do not revoke that access. There are no broad host permissions and no remote code.
What Teams Can Control
- Which AI sites are protected.
- Which risks warn, redact, or block.
- Whether a user can continue after a warning.
- Whether local settings are locked by the organization.
Reporting A Security Issue
Please share only safe reproduction details:
- Browser name and version.
- Extension version.
- Supported AI site name.
- Steps to reproduce using made-up sample text.
- What you expected and what happened.
Do not send real prompts, secrets, customer records, account screenshots, browser history, or payment information.
Trust Status
The product is preparing for paid beta review. It is not a formal compliance certification, a full enterprise security replacement, or a promise of zero false positives.