diff --git a/ietf/person/factories.py b/ietf/person/factories.py
index 5a4d784b7..3db5709ba 100644
--- a/ietf/person/factories.py
+++ b/ietf/person/factories.py
@@ -19,10 +19,12 @@ from ietf.person.name import unidecode_name
 fake = faker.Factory.create()
 
 def random_faker():
-    # The transliteration of arabic names introduces non-alphabetic
-    # characgters that don't work with the draft author extraction code, and
-    # also don't seem to match the way people with arabic names romanize them.
-    locales = set( [ l for l in faker.config.AVAILABLE_LOCALES if not l.startswith('ar_') ] )
+    # The transliteration of some arabic and devanagari names introduces
+    # non-alphabetic characgters that don't work with the draft author
+    # extraction code, and also don't seem to match the way people with arabic
+    # names romanize arabic names.  Exlude those locales from name generation
+    # in order to avoid test failures.
+    locales = set( [ l for l in faker.config.AVAILABLE_LOCALES if not (l.startswith('ar_') or l.startswith('sg_') ] )
     return faker.Faker(random.sample(locales, 1)[0])
 
 class UserFactory(factory.DjangoModelFactory):