diff --git a/django/utils/translation/trans_real.py b/django/utils/translation/trans_real.py
index c1e64d4ebd4a4..46a94b99ffbae 100644
--- a/django/utils/translation/trans_real.py
+++ b/django/utils/translation/trans_real.py
@@ -30,6 +30,11 @@
 # magic gettext number to separate context from message
 CONTEXT_SEPARATOR = "\x04"
 
+# Maximum number of characters that will be parsed from the Accept-Language
+# header to prevent possible denial of service or memory exhaustion attacks.
+# About 10x longer than the longest value shown on MDN’s Accept-Language page.
+ACCEPT_LANGUAGE_HEADER_MAX_LENGTH = 500
+
 # Format of Accept-Language header values. From RFC 9110 Sections 12.4.2 and
 # 12.5.4, and RFC 5646 Section 2.1.
 accept_language_re = _lazy_re_compile(
@@ -582,7 +587,7 @@ def get_language_from_request(request, check_path=False):
 
 
 @functools.lru_cache(maxsize=1000)
-def parse_accept_lang_header(lang_string):
+def _parse_accept_lang_header(lang_string):
     """
     Parse the lang_string, which is the body of an HTTP Accept-Language
     header, and return a tuple of (lang, q-value), ordered by 'q' values.
@@ -604,3 +609,27 @@ def parse_accept_lang_header(lang_string):
         result.append((lang, priority))
     result.sort(key=lambda k: k[1], reverse=True)
     return tuple(result)
+
+
+def parse_accept_lang_header(lang_string):
+    """
+    Parse the value of the Accept-Language header up to a maximum length.
+
+    The value of the header is truncated to a maximum length to avoid potential
+    denial of service and memory exhaustion attacks. Excessive memory could be
+    used if the raw value is very large as it would be cached due to the use of
+    functools.lru_cache() to avoid repetitive parsing of common header values.
+    """
+    # If the header value doesn't exceed the maximum allowed length, parse it.
+    if len(lang_string) <= ACCEPT_LANGUAGE_HEADER_MAX_LENGTH:
+        return _parse_accept_lang_header(lang_string)
+
+    # If there is at least one comma in the value, parse up to the last comma
+    # before the max length, skipping any truncated parts at the end of the
+    # header value.
+    if (index := lang_string.rfind(",", 0, ACCEPT_LANGUAGE_HEADER_MAX_LENGTH)) > 0:
+        return _parse_accept_lang_header(lang_string[:index])
+
+    # Don't attempt to parse if there is only one language-range value which is
+    # longer than the maximum allowed length and so truncated.
+    return ()
diff --git a/docs/releases/3.2.17.txt b/docs/releases/3.2.17.txt
index 9eba24d72f19c..fcc097c5ccdbb 100644
--- a/docs/releases/3.2.17.txt
+++ b/docs/releases/3.2.17.txt
@@ -6,4 +6,12 @@ Django 3.2.17 release notes
 
 Django 3.2.17 fixes a security issue with severity "moderate" in 3.2.16.
 
-...
+CVE-2023-23969: Potential denial-of-service via ``Accept-Language`` headers
+===========================================================================
+
+The parsed values of ``Accept-Language`` headers are cached in order to avoid
+repetitive parsing. This leads to a potential denial-of-service vector via
+excessive memory usage if large header values are sent.
+
+In order to avoid this vulnerability, the ``Accept-Language`` header is now
+parsed up to a maximum length.
diff --git a/docs/releases/4.0.9.txt b/docs/releases/4.0.9.txt
index f05b043c34820..d13f3a2cf6c88 100644
--- a/docs/releases/4.0.9.txt
+++ b/docs/releases/4.0.9.txt
@@ -6,4 +6,12 @@ Django 4.0.9 release notes
 
 Django 4.0.9 fixes a security issue with severity "moderate" in 4.0.8.
 
-...
+CVE-2023-23969: Potential denial-of-service via ``Accept-Language`` headers
+===========================================================================
+
+The parsed values of ``Accept-Language`` headers are cached in order to avoid
+repetitive parsing. This leads to a potential denial-of-service vector via
+excessive memory usage if large header values are sent.
+
+In order to avoid this vulnerability, the ``Accept-Language`` header is now
+parsed up to a maximum length.
diff --git a/docs/releases/4.1.6.txt b/docs/releases/4.1.6.txt
index e97c25aea0b16..04b75683f22c7 100644
--- a/docs/releases/4.1.6.txt
+++ b/docs/releases/4.1.6.txt
@@ -4,8 +4,18 @@ Django 4.1.6 release notes
 
 *February 1, 2023*
 
-Django 4.1.6 fixes a security issue with severity "moderate" and several bugs
-in 4.1.5.
+Django 4.1.6 fixes a security issue with severity "moderate" and a bug in
+4.1.5.
+
+CVE-2023-23969: Potential denial-of-service via ``Accept-Language`` headers
+===========================================================================
+
+The parsed values of ``Accept-Language`` headers are cached in order to avoid
+repetitive parsing. This leads to a potential denial-of-service vector via
+excessive memory usage if large header values are sent.
+
+In order to avoid this vulnerability, the ``Accept-Language`` header is now
+parsed up to a maximum length.
 
 Bugfixes
 ========
diff --git a/tests/i18n/tests.py b/tests/i18n/tests.py
index 0adacb765e4cd..43d1682c96037 100644
--- a/tests/i18n/tests.py
+++ b/tests/i18n/tests.py
@@ -1717,6 +1717,14 @@ def test_parse_spec_http_header(self):
             ("de;q=0.", [("de", 0.0)]),
             ("en; q=1,", [("en", 1.0)]),
             ("en; q=1.0, * ; q=0.5", [("en", 1.0), ("*", 0.5)]),
+            (
+                "en" + "-x" * 20,
+                [("en-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x-x", 1.0)],
+            ),
+            (
+                ", ".join(["en; q=1.0"] * 20),
+                [("en", 1.0)] * 20,
+            ),
             # Bad headers
             ("en-gb;q=1.0000", []),
             ("en;q=0.1234", []),
@@ -1733,6 +1741,10 @@ def test_parse_spec_http_header(self):
             ("", []),
             ("en;q=1e0", []),
             ("en-au;q=１.０", []),
+            # Invalid as language-range value too long.
+            ("xxxxxxxx" + "-xxxxxxxx" * 500, []),
+            # Header value too long, only parse up to limit.
+            (", ".join(["en; q=1.0"] * 500), [("en", 1.0)] * 45),
         ]
         for value, expected in tests:
             with self.subTest(value=value):