diff --git a/litellm/proxy/policy_engine/policy_endpoints.py b/litellm/proxy/policy_engine/policy_endpoints.py
index 3a25e249a4e..35e7f09e806 100644
--- a/litellm/proxy/policy_engine/policy_endpoints.py
+++ b/litellm/proxy/policy_engine/policy_endpoints.py
@@ -4,7 +4,7 @@ CRUD ENDPOINTS FOR POLICIES
 Provides REST API endpoints for managing policies and policy attachments.
 """
 
-from typing import Optional
+from typing import Any, Dict, Optional
 
 from fastapi import APIRouter, Depends, HTTPException
 
@@ -16,6 +16,7 @@ from litellm.proxy.policy_engine.pipeline_executor import PipelineExecutor
 from litellm.proxy.policy_engine.policy_registry import get_policy_registry
 from litellm.types.proxy.policy_engine import (
     GuardrailPipeline,
+    PipelineExecutionResult,
     PipelineTestRequest,
     PolicyAttachmentCreateRequest,
     PolicyAttachmentDBResponse,
@@ -32,6 +33,43 @@ from litellm.types.proxy.policy_engine import (
 
 router = APIRouter()
 
+_PIPELINE_RESPONSE_INTERNAL_KEYS = {
+    "litellm_parent_otel_span",
+    "parent_otel_span",
+    "user_api_key_auth",
+}
+
+
+def _sanitize_pipeline_response_value(value: Any) -> Any:
+    if isinstance(value, dict):
+        return {
+            str(key): _sanitize_pipeline_response_value(item)
+            for key, item in value.items()
+            if str(key) not in _PIPELINE_RESPONSE_INTERNAL_KEYS
+        }
+
+    if isinstance(value, (list, tuple, set)):
+        return [_sanitize_pipeline_response_value(item) for item in value]
+
+    if value is None or isinstance(value, (str, int, float, bool)):
+        return value
+
+    if hasattr(value, "model_dump"):
+        try:
+            return _sanitize_pipeline_response_value(
+                value.model_dump(mode="json", fallback=str)
+            )
+        except Exception:
+            pass
+
+    return str(value)
+
+
+def _dump_pipeline_result_for_response(
+    result: PipelineExecutionResult,
+) -> Dict[str, Any]:
+    return _sanitize_pipeline_response_value(result.model_dump())
+
 
 # ─────────────────────────────────────────────────────────────────────────────
 # Policy CRUD Endpoints
@@ -601,7 +639,7 @@ async def test_pipeline(
             call_type="completion",
             policy_name="test-pipeline",
         )
-        return result.model_dump()
+        return _dump_pipeline_result_for_response(result)
     except Exception as e:
         verbose_proxy_logger.exception(f"Error testing pipeline: {e}")
         raise HTTPException(status_code=500, detail=str(e))
diff --git a/tests/test_litellm/proxy/policy_engine/test_policy_endpoints.py b/tests/test_litellm/proxy/policy_engine/test_policy_endpoints.py
new file mode 100644
index 00000000000..ee4051e866f
--- /dev/null
+++ b/tests/test_litellm/proxy/policy_engine/test_policy_endpoints.py
@@ -0,0 +1,60 @@
+import json
+
+import pytest
+from fastapi.encoders import jsonable_encoder
+
+from litellm.proxy.policy_engine.policy_endpoints import (
+    _dump_pipeline_result_for_response,
+)
+from litellm.types.proxy.policy_engine import (
+    PipelineExecutionResult,
+    PipelineStepResult,
+)
+
+
+class OpaqueSpan:
+    __slots__ = ()
+
+    def __str__(self) -> str:
+        return "opaque-span"
+
+
+def test_pipeline_test_response_removes_internal_non_json_values():
+    opaque_span = OpaqueSpan()
+    result = PipelineExecutionResult(
+        terminal_action="allow",
+        step_results=[
+            PipelineStepResult(
+                guardrail_name="pii",
+                outcome="pass",
+                action_taken="allow",
+                modified_data={
+                    "metadata": {
+                        "parent_otel_span": opaque_span,
+                        "litellm_parent_otel_span": opaque_span,
+                        "safe": "kept",
+                    },
+                    "non_internal_object": opaque_span,
+                },
+            )
+        ],
+        modified_data={
+            "metadata": {
+                "parent_otel_span": opaque_span,
+                "safe": "kept",
+            },
+        },
+    )
+
+    with pytest.raises(ValueError):
+        jsonable_encoder(result.model_dump())
+
+    dumped = _dump_pipeline_result_for_response(result)
+
+    json.dumps(dumped)
+    assert dumped["modified_data"]["metadata"] == {"safe": "kept"}
+    assert dumped["step_results"][0]["modified_data"]["metadata"] == {"safe": "kept"}
+    assert (
+        dumped["step_results"][0]["modified_data"]["non_internal_object"]
+        == "opaque-span"
+    )