diff --git a/litellm/proxy/_lazy_openapi_snapshot.json b/litellm/proxy/_lazy_openapi_snapshot.json
index 3ddf0def821..4cfb2bf8c38 100644
--- a/litellm/proxy/_lazy_openapi_snapshot.json
+++ b/litellm/proxy/_lazy_openapi_snapshot.json
@@ -19621,7 +19621,7 @@
                   }
                 }
               },
-              "description": "\nUnified rate-limit error.\n\nEvery rate-limit condition surfaced by litellm \u2014 whether it originated from\nan upstream LLM provider, a vendor batch endpoint, or one of litellm's own\nproxy-side limiters (parallel-requests, dynamic-rate, batch-rate, budget,\nmax-iterations, etc.) \u2014 is raised as an instance of this class.\n\nThe :attr:`category` attribute lets callers distinguish the source. See\n:class:`RateLimitErrorCategory` for the available values.\n"
+              "description": "\n    Unified rate-limit error.\n\n    Every rate-limit condition surfaced by litellm \u2014 whether it originated from\n    an upstream LLM provider, a vendor batch endpoint, or one of litellm's own\n    proxy-side limiters (parallel-requests, dynamic-rate, batch-rate, budget,\n    max-iterations, etc.) \u2014 is raised as an instance of this class.\n\n    The :attr:`category` attribute lets callers distinguish the source. See\n    :class:`RateLimitErrorCategory` for the available values.\n    "
             },
             "500": {
               "content": {
diff --git a/litellm/proxy/guardrails/guardrail_hooks/azure/base.py b/litellm/proxy/guardrails/guardrail_hooks/azure/base.py
index 2338ed2e30d..d2aa11da7c9 100644
--- a/litellm/proxy/guardrails/guardrail_hooks/azure/base.py
+++ b/litellm/proxy/guardrails/guardrail_hooks/azure/base.py
@@ -21,6 +21,13 @@ AZURE_CONTENT_SAFETY_MAX_TEXT_LENGTH: Final = 10000
 AZURE_CONTENT_SAFETY_TEXT_RECORD_LENGTH: Final = 1000
 
 AZURE_CONTENT_SAFETY_DEFAULT_API_VERSION: Final = "2024-09-01"
+JAVELIN_API_VERSION_STORED_BY_OLDER_RELEASES: Final = "v1"
+
+
+def resolve_content_safety_api_version(configured: str | None) -> str:
+    if not configured or configured == JAVELIN_API_VERSION_STORED_BY_OLDER_RELEASES:
+        return AZURE_CONTENT_SAFETY_DEFAULT_API_VERSION
+    return configured
 
 
 class AzureGuardrailBase:
@@ -58,7 +65,7 @@ class AzureGuardrailBase:
         Returns:
             Parsed JSON response dict.
         """
-        api_version: Final = self.api_version or AZURE_CONTENT_SAFETY_DEFAULT_API_VERSION
+        api_version: Final = resolve_content_safety_api_version(self.api_version)
         url: Final = f"{self.api_base}/contentsafety/{endpoint_path}?api-version={api_version}"
         headers: Final = {
             "Ocp-Apim-Subscription-Key": self.api_key,
diff --git a/tests/test_litellm/proxy/guardrails/guardrail_hooks/azure/test_azure_text_moderation.py b/tests/test_litellm/proxy/guardrails/guardrail_hooks/azure/test_azure_text_moderation.py
index 1798565f383..acb15b4d869 100644
--- a/tests/test_litellm/proxy/guardrails/guardrail_hooks/azure/test_azure_text_moderation.py
+++ b/tests/test_litellm/proxy/guardrails/guardrail_hooks/azure/test_azure_text_moderation.py
@@ -494,3 +494,38 @@ async def test_config_without_api_version_calls_documented_azure_api_version():
     assert mock_post.call_args.kwargs["url"] == (
         "https://example.cognitiveservices.azure.com/contentsafety/text:analyze?api-version=2024-09-01"
     )
+
+
+@pytest.mark.parametrize(
+    ("stored_api_version", "expected_api_version"),
+    [("v1", "2024-09-01"), ("2023-10-01", "2023-10-01")],
+)
+@pytest.mark.asyncio
+async def test_guardrail_loaded_with_stored_api_version_calls_azure_at(stored_api_version, expected_api_version):
+    """Releases before the api_version default fix saved every guardrail created
+    through the API or dashboard with Javelin's "v1", which Azure always answers
+    with 404. A row like that must reach Azure at the documented default, while a
+    real Azure version an admin chose is sent as written."""
+    handler = InMemoryGuardrailHandler()
+    registered = handler.initialize_guardrail(
+        guardrail={
+            "guardrail_name": f"azure-text-moderation-stored-{stored_api_version}",
+            "litellm_params": {
+                "guardrail": "azure/text_moderations",
+                "mode": "pre_call",
+                "api_key": "azure_text_moderation_api_key",
+                "api_base": "https://example.cognitiveservices.azure.com",
+                "api_version": stored_api_version,
+            },
+        }
+    )
+    assert registered is not None
+    guardrail = handler.guardrail_id_to_custom_guardrail[registered["guardrail_id"]]
+
+    with patch.object(guardrail.async_handler, "post", return_value=_moderation_response(0)) as mock_post:
+        result = await guardrail.apply_guardrail(inputs={"texts": ["hello"]}, request_data={}, input_type="request")
+
+    assert result == {"texts": ["hello"]}
+    assert mock_post.call_args.kwargs["url"] == (
+        f"https://example.cognitiveservices.azure.com/contentsafety/text:analyze?api-version={expected_api_version}"
+    )