diff --git a/tests/entrypoints/launchers/api_server/test_api_server_process_manager.py b/tests/entrypoints/launchers/api_server/test_api_server_process_manager.py
index 93d0c49700..40fefc2374 100644
--- a/tests/entrypoints/launchers/api_server/test_api_server_process_manager.py
+++ b/tests/entrypoints/launchers/api_server/test_api_server_process_manager.py
@@ -401,3 +401,52 @@ def test_gather_actual_addresses_child_crash_before_report():
         manager.shutdown()
         time.sleep(0.2)
         sock.close()
+
+
+def test_rust_frontend_launch_log_redacts_credentials(monkeypatch, caplog):
+    """The Rust frontend command carries every non-default arg as JSON.
+    Credentials must not reach the log line."""
+    import subprocess as subprocess_mod
+
+    from vllm.entrypoints.launchers.cli_args import make_arg_parser
+    from vllm.utils.argparse_utils import FlexibleArgumentParser
+    from vllm.v1.utils import RustFrontendProcessManager
+
+    hf_token = "hf_TESTTOKEN123"
+    api_key = "sk-TESTKEY456"
+    args = make_arg_parser(FlexibleArgumentParser()).parse_args(
+        ["--model", "org/model", "--hf-token", hf_token, "--api-key", api_key]
+    )
+
+    class _FakeProc:
+        pid = 4321
+        returncode = 0
+
+        def poll(self):
+            return 0
+
+        def wait(self, timeout=None):
+            return 0
+
+    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    try:
+        monkeypatch.setattr(subprocess_mod, "Popen", lambda *a, **kw: _FakeProc())
+        with caplog.at_level("INFO", logger="vllm.v1.utils"):
+            RustFrontendProcessManager(
+                binary_path="/nonexistent/vllm-rs",
+                sock=sock,
+                args=args,
+                input_address="ipc:///tmp/in",
+                output_address="ipc:///tmp/out",
+                engine_start_index=0,
+                engine_count=1,
+                data_parallel_size=1,
+            )
+    finally:
+        sock.close()
+
+    message = caplog.text
+    assert "Launching Rust frontend:" in message
+    assert hf_token not in message
+    assert api_key not in message
+    assert '"hf_token": "***"' in message
diff --git a/tests/entrypoints/serve/utils/test_api_utils.py b/tests/entrypoints/serve/utils/test_api_utils.py
index 989c8be208..b1bf765595 100644
--- a/tests/entrypoints/serve/utils/test_api_utils.py
+++ b/tests/entrypoints/serve/utils/test_api_utils.py
@@ -8,8 +8,8 @@ import pytest
 from vllm.entrypoints.openai.engine.protocol import StreamOptions
 from vllm.entrypoints.serve.utils import api_utils
 from vllm.entrypoints.serve.utils.api_utils import (
-    _redact_sensitive_args,
     get_max_tokens,
+    redact_sensitive_args,
     should_include_usage,
 )
 
@@ -121,7 +121,7 @@ class TestRedactSensitiveArgs:
 
     def test_redact_replaces_sensitive_values_only(self):
         args = {"api_key": self.API_KEY, "hf_token": "hf_secret", "other": "visible"}
-        redacted = _redact_sensitive_args(args)
+        redacted = redact_sensitive_args(args)
         assert redacted == {
             "api_key": "***",
             "hf_token": "***",
@@ -136,7 +136,7 @@ class TestRedactSensitiveArgs:
 
     def test_no_sensitive_fields_returns_original(self):
         args = {"model_tag": "org/model", "other": "visible"}
-        assert _redact_sensitive_args(args) is args
+        assert redact_sensitive_args(args) is args
 
     def test_api_key_not_in_log(self, monkeypatch, caplog):
         non_default = {
diff --git a/vllm/entrypoints/serve/utils/api_utils.py b/vllm/entrypoints/serve/utils/api_utils.py
index c3c99e5744..87229a4722 100644
--- a/vllm/entrypoints/serve/utils/api_utils.py
+++ b/vllm/entrypoints/serve/utils/api_utils.py
@@ -271,7 +271,7 @@ def jsonify_non_default_args(
 _SENSITIVE_ARG_FIELDS = frozenset({"api_key", "hf_token"})
 
 
-def _redact_sensitive_args(args: dict[str, Any]) -> dict[str, Any]:
+def redact_sensitive_args(args: dict[str, Any]) -> dict[str, Any]:
     """Return a copy of `args` with sensitive values redacted for logging."""
     if not any(key in _SENSITIVE_ARG_FIELDS for key in args):
         return args
@@ -283,7 +283,7 @@ def _redact_sensitive_args(args: dict[str, Any]) -> dict[str, Any]:
 
 def log_non_default_args(args: Namespace | EngineArgs):
     non_default_args = get_non_default_args(args)
-    logger.info("non-default args: %s", _redact_sensitive_args(non_default_args))
+    logger.info("non-default args: %s", redact_sensitive_args(non_default_args))
 
 
 def should_include_usage(
diff --git a/vllm/v1/utils.py b/vllm/v1/utils.py
index 0501c2ae32..b0411dc9e8 100644
--- a/vllm/v1/utils.py
+++ b/vllm/v1/utils.py
@@ -389,7 +389,12 @@ class RustFrontendProcessManager:
         args_json = json.dumps(args_dict, sort_keys=True)
         cmd.extend(["--args-json", args_json])
 
-        logger.info("Launching Rust frontend: %s", " ".join(cmd))
+        # The subprocess needs the real values, but the log must not carry
+        # credentials such as api_key or hf_token.
+        from vllm.entrypoints.serve.utils.api_utils import redact_sensitive_args
+
+        redacted_json = json.dumps(redact_sensitive_args(args_dict), sort_keys=True)
+        logger.info("Launching Rust frontend: %s", " ".join(cmd[:-1] + [redacted_json]))
         self._proc = subprocess.Popen(cmd, pass_fds=(fd,))
 
         # Create a process wrapper with a sentinel fd for monitoring