diff --git a/tests/entrypoints/test_serve.py b/tests/entrypoints/test_serve.py
index 0f51a4320..ad5e33dc1 100644
--- a/tests/entrypoints/test_serve.py
+++ b/tests/entrypoints/test_serve.py
@@ -15,7 +15,10 @@ import pytest
 from pytest_mock import MockerFixture
 from vllm.v1.engine.utils import EngineZmqAddresses
 
-from vllm_omni.config.omni_config import VllmOmniDiffusionStageConfig
+from vllm_omni.config.omni_config import (
+    VllmOmniDiffusionStageConfig,
+    normalize_and_validate_diffusion_engine_ingress_kwargs,
+)
 from vllm_omni.config.resolver import OmniConfigResolution
 from vllm_omni.engine.stage_engine_startup import StageReplicaResources
 from vllm_omni.engine.stage_runtime import StageEngineLaunch
@@ -87,6 +90,26 @@ def _parse_serve_args(argv: list[str]) -> TrackingNamespace:
     return parser.parse_args(argv)
 
 
+def test_no_guardrails_is_only_forwarded_as_model_config(mocker: MockerFixture) -> None:
+    """The CLI alias must not reach the strict diffusion config validator."""
+    parser = TrackingArgumentParser()
+    subparsers = parser.add_subparsers(dest="subparser")
+    OmniServeCommand().subparser_init(subparsers)
+    args = parser.parse_args(["serve", "fake-model", "--omni", "--no-guardrails"])
+    assert args.get_explicit_kwargs_dict()["no_guardrails"] is True
+
+    # Keep the real CLI parser and config validator, but do not start a server.
+    args.headless = True
+    mocker.patch.dict("os.environ", {"VLLM_DISABLE_LOG_LOGO": "1"})
+    mocker.patch("vllm_omni.entrypoints.cli.serve.run_headless")
+    OmniServeCommand.cmd(args)
+
+    explicit = args.get_explicit_kwargs_dict()
+    normalized = normalize_and_validate_diffusion_engine_ingress_kwargs(explicit, stage_id=0)
+    assert normalized["model_config"]["guardrails"] is False
+    assert "no_guardrails" not in explicit
+
+
 def test_omni_serve_requires_model_when_none_provided() -> None:
     """Regression for https://github.com/vllm-project/vllm-omni/issues/4158:
     ``vllm serve --omni`` with no model must fail fast instead of silently
diff --git a/vllm_omni/entrypoints/cli/serve.py b/vllm_omni/entrypoints/cli/serve.py
index 10ae2166d..57514bbe9 100644
--- a/vllm_omni/entrypoints/cli/serve.py
+++ b/vllm_omni/entrypoints/cli/serve.py
@@ -149,7 +149,8 @@ class OmniServeCommand(CLISubcommand):
             args.model_config = model_config
             explicit_keys = getattr(args, "explicit_keys", None)
             if explicit_keys is not None:
-                args.explicit_keys = explicit_keys | {"model_config"}
+                # --no-guardrails is a CLI-only alias, not a diffusion engine arg.
+                args.explicit_keys = (explicit_keys - {"no_guardrails"}) | {"model_config"}
 
         if args.headless:
             run_headless(args)