diff --git a/bleach/html5lib_shim.py b/bleach/html5lib_shim.py
index 20d7e22b..be06aead 100644
--- a/bleach/html5lib_shim.py
+++ b/bleach/html5lib_shim.py
@@ -534,7 +534,18 @@ def next_possible_entity(text):
 
 
 class BleachHTMLSerializer(HTMLSerializer):
-    """HTMLSerializer that undoes & -> & in attributes"""
+    """HTMLSerializer that undoes & -> & in attributes and sets
+    escape_rcdata to True
+    """
+
+    # per the HTMLSerializer.__init__ docstring:
+    #
+    # Whether to escape characters that need to be
+    # escaped within normal elements within rcdata elements such as
+    # style.
+    #
+    escape_rcdata = True
+
     def escape_base_amp(self, stoken):
         """Escapes just bare & in HTML attribute values"""
         # First, undo escaping of &. We need to do this because html5lib's
diff --git a/tests/test_clean.py b/tests/test_clean.py
index 8f64beb2..133cd822 100644
--- a/tests/test_clean.py
+++ b/tests/test_clean.py
@@ -7,7 +7,7 @@
 from bleach import clean
 from bleach.html5lib_shim import Filter
 from bleach.sanitizer import Cleaner
-
+from bleach._vendor.html5lib.constants import rcdataElements
 
 def test_clean_idempotent():
     """Make sure that applying the filter twice doesn't change anything."""
@@ -789,7 +789,7 @@ def test_nonexistent_namespace():
         (
             raw_tag,
             "<noscript><%s></noscript><img src=x onerror=alert(1) />" % raw_tag,
-            "<noscript><%s></noscript>&lt;img src=x onerror=alert(1) /&gt;" % raw_tag,
+            "<noscript>&lt;%s&gt;</noscript>&lt;img src=x onerror=alert(1) /&gt;" % raw_tag,
         )
         for raw_tag in _raw_tags
     ],
@@ -799,6 +799,29 @@ def test_noscript_rawtag_(raw_tag, data, expected):
     assert clean(data, tags=["noscript", raw_tag]) == expected
 
 
+@pytest.mark.parametrize(
+    "namespace_tag, rc_data_element_tag, data, expected",
+    [
+        (
+            namespace_tag,
+            rc_data_element_tag,
+            "<%s><%s><img src=x onerror=alert(1)>" % (namespace_tag, rc_data_element_tag),
+            "<%s><%s>&lt;img src=x onerror=alert(1)&gt;</%s></%s>" % (namespace_tag, rc_data_element_tag, rc_data_element_tag, namespace_tag),
+        )
+        for namespace_tag in ["math", "svg"]
+        # https://dev.w3.org/html5/html-author/#rcdata-elements
+        # https://html.spec.whatwg.org/index.html#parsing-html-fragments
+        # in html5lib: 'style', 'script', 'xmp', 'iframe', 'noembed', 'noframes', and 'noscript'
+        for rc_data_element_tag in rcdataElements
+    ],
+)
+def test_namespace_rc_data_element_strip_false(namespace_tag, rc_data_element_tag, data, expected):
+    # refs: bug 1621692 / GHSA-m6xf-fq7q-8743
+    #
+    # browsers will pull the img out of the namespace and rc data tag resulting in XSS
+    assert clean(data, tags=[namespace_tag, rc_data_element_tag], strip=False) == expected
+
+
 def get_ids_and_tests():
     """Retrieves regression tests from data/ directory