privacy.html

<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1" />
  <title>Privacy Policy | Clueside</title>
  <style>
    body {
      margin: 0 auto;
      max-width: 40rem;
      padding: 2rem 1.25rem 4rem;
      font-family: system-ui, -apple-system, Segoe UI, Roboto, sans-serif;
      line-height: 1.5;
      color: #1a1a1a;
    }
    h1 { font-size: 1.6rem; margin-bottom: 0.25rem; }
    h2 { font-size: 1.15rem; margin-top: 1.75rem; }
    .meta { color: #555; font-size: 0.9rem; }
    table { border-collapse: collapse; width: 100%; font-size: 0.95rem; }
    th, td { border: 1px solid #ddd; padding: 0.4rem 0.5rem; text-align: left; vertical-align: top; }
    th { background: #f4f4f4; }
  </style>
</head>
<body>
  <h1>Privacy Policy</h1>
  <p class="meta">
    Clueside<br />
    Author: Luis Felipe Abarca<br />
    License: MIT<br />
    Last updated: August 2026
  </p>

  <h2>Summary</h2>
  <p>
    This Chrome extension analyzes images entirely on your device. Image bytes are not uploaded
    to any server for inference, storage, or analytics.
  </p>

  <h2>What stays on your device</h2>
  <ul>
    <li>Webpage images fetched so pixels can be decoded locally</li>
    <li>Files you drop into the toolbar popup</li>
    <li>Model inference (WebGPU or WASM in an offscreen document)</li>
    <li>Metadata parsing (C2PA, EXIF, embedded text)</li>
    <li>Settings (threshold, enable, auto-scan) in <code>chrome.storage.sync</code></li>
  </ul>

  <h2>What may use the network</h2>
  <ul>
    <li>
      <strong>Packaged models:</strong> both model weights (CommunityForensics ONNX and
      DINOv2-small ONNX with its probe head) ship inside the extension package. The weights are
      downloaded from Hugging Face at build/packaging time (<code>npm run fetch-model</code>),
      never by the installed extension. Models and inference assets are never fetched at
      runtime.
    </li>
    <li>
      <strong>Update check:</strong> at most once a day the background worker asks the GitHub
      releases API for the latest version string so the popup can show an update banner. No
      image data or browsing data is sent; offline it fails silently.
    </li>
    <li>
      Ordinary page loads still use the site's own network traffic. This extension does not add
      cloud inference calls and does not send image pixels to a backend.
    </li>
  </ul>

  <h2>Permissions</h2>
  <table>
    <thead>
      <tr><th>Permission</th><th>Why</th></tr>
    </thead>
    <tbody>
      <tr>
        <td>host_permissions / all_urls</td>
        <td>Fetch page image URLs so pixels can be decoded without canvas taint. Local analysis only.</td>
      </tr>
      <tr>
        <td>offscreen</td>
        <td>Run ONNX inference in a document with canvas, WebGPU, or WASM.</td>
      </tr>
      <tr>
        <td>storage</td>
        <td>Save your threshold and toggles.</td>
      </tr>
      <tr>
        <td>alarms</td>
        <td>Schedule the optional daily GitHub release-version check.</td>
      </tr>
    </tbody>
  </table>

  <h2>Data we do not collect</h2>
  <ul>
    <li>No analytics</li>
    <li>No image uploads</li>
    <li>No account system</li>
    <li>No external inference APIs</li>
  </ul>

  <h2>Children</h2>
  <p>This extension does not knowingly collect personal information from anyone, including children.</p>

  <h2>Changes</h2>
  <p>If this policy changes, the date at the top of this page will be updated.</p>

  <h2>Contact</h2>
  <p>
    Source and issues: the project GitHub repository listed with the extension.
    Author: Luis Felipe Abarca. License: MIT.
  </p>
</body>
</html>