diff --git a/changedetectionio/blueprint/tags/templates/groups-overview.html b/changedetectionio/blueprint/tags/templates/groups-overview.html
index 31b040344..dbfdb12f4 100644
--- a/changedetectionio/blueprint/tags/templates/groups-overview.html
+++ b/changedetectionio/blueprint/tags/templates/groups-overview.html
@@ -30,13 +30,13 @@ html[data-darkmode="true"] .watch-tag-list.tag-{{ class_name }} {
     <form class="pure-form" action="{{ url_for('tags.form_tag_add') }}" method="POST" id="new-watch-form">
         <input type="hidden" name="csrf_token" value="{{ csrf_token() }}" >
         <fieldset>
-            <legend>{{ _('Add a new organisational tag') }}</legend>
+            <legend>{{ _('Add a new organisational tag / group name.') }}</legend>
             <div id="watch-add-wrapper-zone">
                 <div>
                     {{ render_simple_field(form.name, placeholder=_("Watch group / tag")) }}
                 </div>
                 <div>
-                    {{ render_simple_field(form.save_button, title=_("Save") ) }}
+                    {{ render_simple_field(form.save_button, title=_("Save"), label=false ) }}
                 </div>
             </div>
             <br>
diff --git a/changedetectionio/flask_app.py b/changedetectionio/flask_app.py
index 8454249a7..7a0afd374 100644
--- a/changedetectionio/flask_app.py
+++ b/changedetectionio/flask_app.py
@@ -671,27 +671,18 @@ def _jinja2_filter_fetcher_status_icons(fetcher_name):
 
     return ''
 
-
-_RE_SANITIZE_TAG = re.compile(r'[^a-zA-Z0-9]')
-
-
 @app.template_filter('sanitize_tag_class')
 def _jinja2_filter_sanitize_tag_class(tag_title):
     """Sanitize a tag title to create a valid CSS class name.
-    Removes all non-alphanumeric characters and converts to lowercase.
-
     Args:
         tag_title: The tag title string
 
     Returns:
         str: A sanitized string suitable for use as a CSS class name
     """
-    # Remove all non-alphanumeric characters and convert to lowercase
-    sanitized = _RE_SANITIZE_TAG.sub('', tag_title).lower()
-    # Ensure it starts with a letter (CSS requirement)
-    if sanitized and not sanitized[0].isalpha():
-        sanitized = 'tag' + sanitized
-    return sanitized if sanitized else 'tag'
+    #
+    tag_class_name = hashlib.sha256(tag_title.encode('utf-8')).hexdigest()[:16]
+    return tag_class_name if tag_class_name else 'tag'
 
 
 # Import login_optionally_required from auth_decorator
diff --git a/changedetectionio/templates/_helpers.html b/changedetectionio/templates/_helpers.html
index de1cd8bdf..87ca1a9dc 100644
--- a/changedetectionio/templates/_helpers.html
+++ b/changedetectionio/templates/_helpers.html
@@ -74,9 +74,13 @@
   </div>
 {% endmacro %}
 
-
+{# label - false - Dont render, None - use default field.label.text #}
 {% macro render_simple_field(field, label=None) %}
-  <span class="label {% if field.errors %}error{% endif %}"><label for="{{ field.id }}">{{ (label if label is not none else field.label.text) | string | forceescape }}</label></span>
+  {%- if label is not false -%}
+  <span class="label {% if field.errors %}error{% endif %}">
+      <label for="{{ field.id }}">{{ (label if label is not none else field.label.text) | string | forceescape }}</label>
+  </span>
+  {%- endif -%}
   <span {% if field.errors %} class="error" {% endif %}>{{ field(**kwargs)|safe }}
   {% if field.errors %}
     <ul class=errors>
diff --git a/changedetectionio/tests/unit/test_sanitize_tag_class.py b/changedetectionio/tests/unit/test_sanitize_tag_class.py
new file mode 100644
index 000000000..9948e5d3a
--- /dev/null
+++ b/changedetectionio/tests/unit/test_sanitize_tag_class.py
@@ -0,0 +1,93 @@
+#!/usr/bin/env python3
+
+"""A group/tag title must map to its own CSS class, whatever alphabet it is written in.
+
+The auto colour for a group is emitted as a stylesheet rule keyed on the class name that
+`sanitize_tag_class` derives from the title (see groups-overview.html / watch-overview.html).
+
+The original filter built that class by stripping every non-alphanumeric character, which for
+an ASCII title was fine but for a title written entirely in Japanese, Chinese, Korean, Cyrillic
+(etc.) stripped the whole string - so every one of those groups fell through to the literal
+`tag-tag` class, shared one stylesheet rule, and rendered in the same colour (#4451).
+
+Hashing the title instead keeps the output in [0-9a-f] (always a valid class name) while giving
+every distinct title a distinct class.
+
+run from dir above changedetectionio/ dir
+python3 -m unittest changedetectionio.tests.unit.test_sanitize_tag_class
+"""
+
+import re
+import unittest
+
+from changedetectionio.flask_app import _jinja2_filter_sanitize_tag_class as sanitize_tag_class
+
+# The class is used as `tag-{{ class_name }}`, so it may only contain characters that are legal
+# in a CSS identifier - and must never be empty.
+CSS_SAFE = re.compile(r'^[0-9a-f]+$')
+
+
+class TestSanitizeTagClass(unittest.TestCase):
+
+    def test_output_is_always_a_safe_css_identifier(self):
+        """Whatever goes in, what comes out must be usable as `.tag-<name>`"""
+        titles = [
+            'Price drops',
+            'price drops',
+            '価格変動',
+            '가격 변동',
+            '价格变动',
+            'Цена',
+            'Ελλάδα',
+            '🚨 alerts',
+            'a b\tc\nd',
+            'has "quotes" and {braces}',
+            '../../etc/passwd',
+            '} .foo { background-color: red;',
+            '-',
+            '0',
+            ' ',
+            '',
+        ]
+        for title in titles:
+            with self.subTest(title=title):
+                class_name = sanitize_tag_class(title)
+                self.assertTrue(class_name, f"{title!r} must not produce an empty class name")
+                self.assertRegex(class_name, CSS_SAFE,
+                                 f"{title!r} produced a class name that would break the stylesheet")
+
+    def test_non_ascii_titles_get_their_own_class(self):
+        """#4451 - these all collapsed to the same class, so they all got the same auto colour"""
+        titles = ['価格変動', '在庫状況', '空室情報']
+        class_names = [sanitize_tag_class(t) for t in titles]
+
+        self.assertEqual(len(set(class_names)), len(titles),
+                         f"non-ASCII group titles must not share a class name, got {class_names}")
+        for class_name in class_names:
+            self.assertNotEqual(class_name, 'tag',
+                                "non-ASCII titles must not fall back to the shared 'tag' class")
+
+    def test_titles_that_differ_only_in_punctuation_get_their_own_class(self):
+        """Stripping non-alphanumerics also merged these; the auto colour must still differ"""
+        self.assertNotEqual(sanitize_tag_class('re-stock'), sanitize_tag_class('restock'))
+        self.assertNotEqual(sanitize_tag_class('My Tag'), sanitize_tag_class('my tag'))
+
+    def test_is_stable(self):
+        """The rule and the element that uses it are rendered by separate calls - they must agree"""
+        for title in ['Price drops', '価格変動', '']:
+            with self.subTest(title=title):
+                self.assertEqual(sanitize_tag_class(title), sanitize_tag_class(title))
+
+    def test_is_deterministic_across_runs(self):
+        """Not salted/randomised per process, so a class name can be relied on in a test or CSS"""
+        self.assertEqual(sanitize_tag_class('Price drops'), '03d0b363b6f919cc')
+        self.assertEqual(sanitize_tag_class('価格変動'), 'd42fc3593d9c3c5a')
+
+    def test_many_titles_stay_unique(self):
+        titles = [f"group {i}" for i in range(500)] + [f"グループ{i}" for i in range(500)]
+        class_names = {sanitize_tag_class(t) for t in titles}
+        self.assertEqual(len(class_names), len(titles), "class names must not collide")
+
+
+if __name__ == '__main__':
+    unittest.main()