A secure rich text editor is a shared job.
The editor ships the controls. Your server finishes them. Here is the split, area by area.
Details
A rich text field accepts HTML from users, uploads files and may call an AI provider. Each of those is a way in. No editor makes an application secure by itself, so this page sets out what RichTextEditor does, what your application must do, and where to check each claim.
Your servers, not ours.
What the editor does, and what you do
Stored XSS
In the editor: Sanitizes HTML in the browser, including pasted content, with bounded paste processing.
Your responsibility: Sanitize again on the server before you store, because a request can skip the browser.
Content Security Policy
In the editor: Runs under a strict policy; the CSP guide lists the settings it needs.
Your responsibility: Choose and send the policy, and the trusted origins.
Uploads
In the editor: Reference handlers check size, extension and file signature (magic bytes).
Your responsibility: Authorize each file, scan for malware, and decide retention and private storage.
AI
In the editor: Provider keys stay on your server behind a resolver, with size and rate controls and audit hooks.
Your responsibility: Pick the provider and region, store the key in your vault, and decide who may use AI.
Collaboration
In the editor: Room-scoped signed tokens, an origin allowlist, and payload and room limits.
Your responsibility: Terminate TLS, protect the secret, and mint tokens only for users who may open the document.
Telemetry
In the editor: No required RichScripts telemetry endpoint; files are served from your origin.
Your responsibility: Check the network panel yourself, and block other origins with your CSP.
Check it yourself
- XSS: the content security demo runs real attack payloads and shows what the sanitizer removes. The stored XSS article covers the server side.
- CSP: Content-Security-Policy for rich text editors gives a policy to start from and the settings the editor needs.
- Uploads: store images as files shows how to turn base64 images off and refuse them on the server.
- AI keys: streaming AI through your server explains why the browser should never hold a provider key.
- Outbound requests: does your editor phone home? walks through the network-panel check.
What is not available
- No SOC 2 report, penetration-test attestation or ISO certification is published today. The security architecture overview is a description of the product, not an audit.
- Self-hosting moves operations to you. Patching, TLS, secrets and backups for the services you run are your responsibility. See the self-hosted editor page.
- Report a vulnerability privately through support, with the version, steps to reproduce and impact. Do not send production secrets or customer content.