diff --git a/litellm/proxy/spend_tracking/spend_tracking_utils.py b/litellm/proxy/spend_tracking/spend_tracking_utils.py
index 8a963ec0134..e9a079d622b 100644
--- a/litellm/proxy/spend_tracking/spend_tracking_utils.py
+++ b/litellm/proxy/spend_tracking/spend_tracking_utils.py
@@ -137,6 +137,17 @@ def _get_spend_logs_metadata(
     clean_metadata["litellm_overhead_time_ms"] = litellm_overhead_time_ms
     clean_metadata["cost_breakdown"] = cost_breakdown
 
+    # Redact error_message when prompt/response storage is disabled.
+    # str(exception) for provider errors (e.g. BadRequestError) includes the raw
+    # response body (Received={...}), which violates the no-content-logging policy.
+    if not _should_store_prompts_and_responses_in_spend_logs():
+        error_info = clean_metadata.get("error_information")
+        if isinstance(error_info, dict) and error_info.get("error_message"):
+            clean_metadata["error_information"] = {
+                **error_info,
+                "error_message": None,
+            }
+
     return clean_metadata
 
 
diff --git a/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py b/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py
index 30b952cd421..88902871372 100644
--- a/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py
+++ b/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py
@@ -1469,3 +1469,61 @@ class TestIsMasterKey:
         master = "sk-master-key-123"
         hashed = hash_token(master)
         assert _is_master_key(api_key=hashed, _master_key=master) is True
+
+
+# ---------------------------------------------------------------------------
+# error_message redaction in spend logs
+# ---------------------------------------------------------------------------
+
+STORE_PROMPTS_PATH = "litellm.proxy.spend_tracking.spend_tracking_utils._should_store_prompts_and_responses_in_spend_logs"
+
+
+def _failure_metadata_with_error(error_msg: str) -> dict:
+    return {
+        "user_api_key": "sk-test",
+        "status": "failure",
+        "error_information": {
+            "error_code": "400",
+            "error_class": "BadRequestError",
+            "llm_provider": "vertex_ai",
+            "traceback": "Traceback ...",
+            "error_message": error_msg,
+        },
+    }
+
+
+@patch(STORE_PROMPTS_PATH, return_value=False)
+def test_error_message_redacted_when_store_prompts_disabled(mock_store):
+    """error_message must be None when store_prompts_in_spend_logs is False."""
+    raw = 'litellm.BadRequestError: Received={"candidates": [{"content": {"role": "model", "parts": [{"text": "secret content"}]}}]}'
+    result = _get_spend_logs_metadata(_failure_metadata_with_error(raw))
+    assert result["error_information"]["error_message"] is None
+
+
+@patch(STORE_PROMPTS_PATH, return_value=True)
+def test_error_message_kept_when_store_prompts_enabled(mock_store):
+    """error_message must be preserved when store_prompts_in_spend_logs is True."""
+    raw = 'litellm.BadRequestError: Received={"candidates": [{"content": {"role": "model"}}]}'
+    result = _get_spend_logs_metadata(_failure_metadata_with_error(raw))
+    assert result["error_information"]["error_message"] == raw
+
+
+@patch(STORE_PROMPTS_PATH, return_value=False)
+def test_other_error_fields_preserved_when_message_redacted(mock_store):
+    """error_code, error_class, llm_provider, and traceback must survive redaction."""
+    raw = "error with content"
+    result = _get_spend_logs_metadata(_failure_metadata_with_error(raw))
+    info = result["error_information"]
+    assert info["error_code"] == "400"
+    assert info["error_class"] == "BadRequestError"
+    assert info["llm_provider"] == "vertex_ai"
+    assert info["traceback"] == "Traceback ..."
+    assert info["error_message"] is None
+
+
+@patch(STORE_PROMPTS_PATH, return_value=False)
+def test_no_error_when_error_information_is_none(mock_store):
+    """No crash when error_information is absent (success path)."""
+    metadata = {"user_api_key": "sk-test", "status": "success"}
+    result = _get_spend_logs_metadata(metadata)
+    assert result.get("error_information") is None