diff --git a/ddtrace/contrib/internal/openai/utils.py b/ddtrace/contrib/internal/openai/utils.py
index f5dfc10efe..0217b1e61d 100644
--- a/ddtrace/contrib/internal/openai/utils.py
+++ b/ddtrace/contrib/internal/openai/utils.py
@@ -89,7 +89,10 @@ def _extract_token_chunk(self, chunk):
         """Attempt to extract the token chunk (last chunk in the stream) from the streamed response."""
         if not self._dd_span._get_ctx_item("_dd.auto_extract_token_chunk"):
             return
-        choice = getattr(chunk, "choices", [None])[0]
+        choices = getattr(chunk, "choices")
+        if not choices:
+            return
+        choice = choices[0]
         if not getattr(choice, "finish_reason", None):
             # Only the second-last chunk in the stream with token usage enabled will have finish_reason set
             return
@@ -152,7 +155,10 @@ async def _extract_token_chunk(self, chunk):
         """Attempt to extract the token chunk (last chunk in the stream) from the streamed response."""
         if not self._dd_span._get_ctx_item("_dd.auto_extract_token_chunk"):
             return
-        choice = getattr(chunk, "choices", [None])[0]
+        choices = getattr(chunk, "choices")
+        if not choices:
+            return
+        choice = choices[0]
         if not getattr(choice, "finish_reason", None):
             return
         try:
diff --git a/releasenotes/notes/fix-token-extraction-0133808742374ef4.yaml b/releasenotes/notes/fix-token-extraction-0133808742374ef4.yaml
new file mode 100644
index 0000000000..cc8c1aa127
--- /dev/null
+++ b/releasenotes/notes/fix-token-extraction-0133808742374ef4.yaml
@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    LLM Observability: This fix resolves an issue where extracting token metadata from openai streamed chat completion token chunks caused an IndexError.