diff --git a/tests/test_main.py b/tests/test_main.py
index 476b77b..271be33 100644
--- a/tests/test_main.py
+++ b/tests/test_main.py
@@ -127,6 +127,8 @@ class TestMain:
                     ),
                 )
             ),
+            # Malformed MIME type
+            (b"...", (b"javascript charset=UTF-8",), True, False, None, None, b"text/plain"),
         ],
     )
     def test_extract_mime(
diff --git a/xtractmime/__init__.py b/xtractmime/__init__.py
index d68247c..9de8882 100644
--- a/xtractmime/__init__.py
+++ b/xtractmime/__init__.py
@@ -1,4 +1,5 @@
 __version__ = "0.2.0"
+import re
 from typing import Optional, Set, Tuple
 from xtractmime._patterns import _APACHE_TYPES, BINARY_BYTES, WHITESPACE_BYTES
 from xtractmime._utils import (
@@ -187,6 +188,27 @@ def _sniff_mislabled_feed(input_bytes: bytes, supplied_type: bytes) -> Optional[
     return supplied_type
 
 
+_TOKEN = br"^\s*[-!#$%&'*+.0-9A-Z^_`a-z{|}~]+\s*$"
+
+
+def _is_valid_mime_type(mime_type):
+    """Return True if the specified MIME type is valid as per RFC 2045, or
+    False otherwise.
+
+    Only the type and subtype are validated, parameters are ignored.
+    """
+    parts = mime_type.split(b"/", maxsplit=1)
+    if len(parts) < 2:
+        return False
+    _type, subtype_and_params = parts
+    if not re.match(_TOKEN, _type):
+        return False
+    subtype = subtype_and_params.split(b";", maxsplit=1)[0]
+    if not re.match(_TOKEN, subtype):
+        return False
+    return True
+
+
 def extract_mime(
     body: bytes,
     *,
@@ -199,6 +221,8 @@ def extract_mime(
     extra_types = extra_types or tuple()
     supplied_type = content_types[-1] if content_types else b""
     check_for_apache = http_origin and supplied_type in _APACHE_TYPES
+    if not _is_valid_mime_type(supplied_type):
+        supplied_type = b""
     supplied_type = supplied_type.split(b";")[0].strip().lower()
     resource_header = memoryview(body)[:RESOURCE_HEADER_BUFFER_LENGTH]