Hi Gemini Web & Infrastructure Team,
I am reporting a critical frontend issue where the web UI renderer crashes mid-response whenever the model generates code blocks (JSX/TSX, React components, template literals, or standard web syntax).
Symptoms
-
The output streams normally until it reaches a JSX tag or code block hydration point (e.g., right at
return (or<div). -
At that exact moment, the code block fails to render, spilling raw text into the main DOM layout.
-
UI elements (copy buttons, feedback thumbs, syntax highlighting) completely break or freeze.
Root Cause & Console Diagnostic
Inspecting the Chrome Developer Console reveals that this is not a model or Markdown parser bug, but a Content Security Policy (CSP) misconfiguration on the backend.
The server handling my session sends a Content-Security-Policy HTTP header that omits https://*.static.usercontent.goog from its script-src directive. When the frontend attempts to dynamically import code block hydration modules, Chrome blocks them.
Console Logs:
Refused to load the script '[https://njl0.static.usercontent.goog/boq-bard-web/islands/assets/island-bootstrap-runtime-*.js](https://njl0.static.usercontent.goog/boq-bard-web/islands/assets/island-bootstrap-runtime-*.js)'
because it violates the following Content Security Policy directive: "script-src 'unsafe-inline' 'unsafe-eval' blob: data: [https://www.gstatic.com](https://www.gstatic.com) ... [missing *.static.usercontent.goog]"
Verification & Key Findings
-
Staged Rollout Issue: A secondary paid account under the exact same Google Workspace enterprise domain works perfectly. It hits a different backend cluster whose CSP response header explicitly includes
https://*.static.usercontent.goog. -
Client-Side Unfixable: Chrome’s Manifest V3 security policy strictly enforces CSP on primary Google domains at the network level, preventing local browser extensions or header-modifying tools from overriding it.
Requested Action
Could the frontend/infrastructure team please update the Content-Security-Policy header configuration across all active production and staged rollout backend clusters for Gemini Web?
Thank you!