diff --git a/hail/python/hail/backend/service_backend.py b/hail/python/hail/backend/service_backend.py
index 1a2c33a51f2..46767eafbd9 100644
--- a/hail/python/hail/backend/service_backend.py
+++ b/hail/python/hail/backend/service_backend.py
@@ -727,7 +727,7 @@ def set_flags(self, **flags: str):
         self.flags.update(flags)
 
     def get_flags(self, *flags) -> Mapping[str, str]:
-        return frozendict(self.flags)
+        return {flag: self.flags[flag] for flag in flags}
 
     @property
     def requires_lowering(self):
diff --git a/hail/python/test/hail/test_context.py b/hail/python/test/hail/test_context.py
index b2645ae9405..6e1cd312b5c 100644
--- a/hail/python/test/hail/test_context.py
+++ b/hail/python/test/hail/test_context.py
@@ -29,3 +29,7 @@ def test_top_level_functions_are_do_not_error(self):
 
     def test_tmpdir_runs(self):
         isinstance(hl.tmp_dir(), str)
+
+    def test_get_flags(self):
+        assert hl._get_flags() == {}
+        assert list(hl._get_flags('use_new_shuffle')) == 'use_new_shuffle'