diff --git a/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/requirements.oldest.txt b/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/requirements.oldest.txt
index 7b1e5347..1d2c5e62 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/requirements.oldest.txt
+++ b/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/requirements.oldest.txt
@@ -20,6 +20,12 @@
 # pyproject floors by UV_RESOLUTION=lowest-direct on the oldest tox factor, so they are NOT pinned
 # here to avoid drift between the declared bound and the tested version.
 
+# Exception: this change depends on the fully-qualified error.type behavior added in
+# opentelemetry-util-genai 1.1b0, which is not on PyPI yet. Install it from the workspace so the
+# oldest env exercises the required behavior. Remove this once 1.1b0 is released and bump the
+# pyproject floor to >= 1.1b0.
+-e ./util/opentelemetry-util-genai
+
 langchain-openai==0.2.0
 langchain-aws==0.2.2
 langchain-google-genai==2.0.0
diff --git a/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/test_llm_call.py b/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/test_llm_call.py
index ebd585dc..be669736 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/test_llm_call.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-langchain/tests/test_llm_call.py
@@ -329,7 +329,9 @@ def assert_openai_completion_attributes_with_error(
     assert span is not None
     assert span.name == "chat gpt-3.5-turbo"
     attributes = span.attributes
-    assert attributes[error_attributes.ERROR_TYPE] == "AuthenticationError"
+    assert (
+        attributes[error_attributes.ERROR_TYPE] == "openai.AuthenticationError"
+    )
     assert attributes[gen_ai_attributes.GEN_AI_OPERATION_NAME] == "chat"
     assert (
         attributes[gen_ai_attributes.GEN_AI_REQUEST_MODEL] == "gpt-3.5-turbo"
@@ -490,7 +492,9 @@ def assert_duration_metric_when_error(metric, parent_span):
 
 def assert_duration_metric_attributes_when_error(attributes, parent_span):
     assert len(attributes) == 4
-    assert attributes[error_attributes.ERROR_TYPE] == "AuthenticationError"
+    assert (
+        attributes[error_attributes.ERROR_TYPE] == "openai.AuthenticationError"
+    )
     assert attributes.get(gen_ai_attributes.GEN_AI_PROVIDER_NAME) == "openai"
     assert (
         attributes.get(gen_ai_attributes.GEN_AI_OPERATION_NAME)
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/.changelog/282.fixed b/instrumentation/opentelemetry-instrumentation-genai-openai/.changelog/282.fixed
new file mode 100644
index 00000000..ad0e8d4a
--- /dev/null
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/.changelog/282.fixed
@@ -0,0 +1 @@
+Responses that fail now record the actual error type (`error` event code or the response's error code) instead of a generic `RuntimeError`, for both streaming and non-streaming calls. Responses that finish as incomplete (`max_output_tokens`/`content_filter`) are recorded as a finish reason rather than an error.
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch.py b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch.py
index b38b3214..b0622561 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch.py
@@ -17,9 +17,6 @@
     EmbeddingInvocation,
     InferenceInvocation,
 )
-from opentelemetry.util.genai.types import (
-    Error,
-)
 
 from .chat_wrappers import AsyncChatStreamWrapper, ChatStreamWrapper
 from .utils import (
@@ -89,7 +86,7 @@ def traced_method(wrapped, instance, args, kwargs):
             chat_invocation.stop()
             return result
         except Exception as error:
-            chat_invocation.fail(Error(type=type(error), message=str(error)))
+            chat_invocation.fail(error)
             raise
 
     return traced_method
@@ -125,7 +122,7 @@ async def traced_method(wrapped, instance, args, kwargs):
             return result
 
         except Exception as error:
-            chat_invocation.fail(Error(type=type(error), message=str(error)))
+            chat_invocation.fail(error)
             raise
 
     return traced_method
@@ -140,7 +137,7 @@ def traced_method(wrapped, instance, args, kwargs):
         try:
             result = wrapped(*args, **kwargs)
         except Exception as error:
-            invocation.fail(Error(type=type(error), message=str(error)))
+            invocation.fail(error)
             raise
 
         _safe_set_embeddings_response_properties(invocation, result)
@@ -159,7 +156,7 @@ async def traced_method(wrapped, instance, args, kwargs):
         try:
             result = await wrapped(*args, **kwargs)
         except Exception as error:
-            invocation.fail(Error(type=type(error), message=str(error)))
+            invocation.fail(error)
             raise
 
         _safe_set_embeddings_response_properties(invocation, result)
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch_responses.py b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch_responses.py
index 3eb4c443..704b66c5 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch_responses.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/patch_responses.py
@@ -11,6 +11,7 @@
     apply_request_attributes,
     extract_params,
     get_inference_creation_kwargs,
+    get_response_error,
     set_invocation_response_attributes,
 )
 from .response_wrappers import (
@@ -115,12 +116,15 @@ def traced_method(
                     capture_content,
                 )
 
+            response = cast("ResponseResult", parsed_result)
             set_invocation_response_attributes(
-                invocation,
-                cast("ResponseResult", parsed_result),
-                capture_content,
+                invocation, response, capture_content
             )
-            invocation.stop()
+            error = get_response_error(response)
+            if error is not None:
+                invocation.fail(error)
+            else:
+                invocation.stop()
             return result
         except Exception as error:
             invocation.fail(error)
@@ -194,12 +198,15 @@ async def traced_method(
                     capture_content,
                 )
 
+            response = cast("ResponseResult", parsed_result)
             set_invocation_response_attributes(
-                invocation,
-                cast("ResponseResult", parsed_result),
-                capture_content,
+                invocation, response, capture_content
             )
-            invocation.stop()
+            error = get_response_error(response)
+            if error is not None:
+                invocation.fail(error)
+            else:
+                invocation.stop()
             return result
         except Exception as error:
             invocation.fail(error)
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_extractors.py b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_extractors.py
index 3b9bc7b9..0642e8d6 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_extractors.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_extractors.py
@@ -25,6 +25,7 @@
     from openai.types.responses.response_usage import ResponseUsage
 
     from opentelemetry.util.genai.types import (
+        Error,
         InputMessage,
         OutputMessage,
         Text,
@@ -57,6 +58,7 @@
 
 try:
     from opentelemetry.util.genai.types import (
+        Error,
         InputMessage,
         OutputMessage,
         Reasoning,
@@ -66,6 +68,7 @@
         ToolCallRequest as ToolCall,
     )
 except ImportError:
+    Error = None
     InputMessage = None
     OutputMessage = None
     Reasoning = None
@@ -365,6 +368,21 @@ def extract_finish_reasons(response: "Response | None") -> list[str]:
     return list(dict.fromkeys(finish_reasons))
 
 
+def get_response_error(response: "Response | None") -> "Error | None":
+    """Return an ``Error`` when the response failed, else ``None``.
+
+    A failed response carries a ``ResponseError`` (``code`` + ``message``).
+    Incomplete responses (``incomplete_details``) are *not* errors — they
+    surface as a finish reason instead.
+    """
+    if Response is None or Error is None or not isinstance(response, Response):
+        return None
+    error = response.error
+    if error is None or not getattr(error, "code", None):
+        return None
+    return Error(type=error.code, message=getattr(error, "message", None))
+
+
 def get_inference_creation_kwargs(
     params: ResponseRequestParams,
     client_instance: object,
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_wrappers.py b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_wrappers.py
index 8a3f0f11..d90543e2 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_wrappers.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/src/opentelemetry/instrumentation/genai/openai/response_wrappers.py
@@ -17,9 +17,11 @@
 
 try:
     from opentelemetry.instrumentation.genai.openai.response_extractors import (  # pylint: disable=no-name-in-module
+        get_response_error,
         set_invocation_response_attributes,
     )
 except ImportError:
+    get_response_error = None
     set_invocation_response_attributes = None
 
 if TYPE_CHECKING:
@@ -140,10 +142,10 @@ def _stop(
         self._self_invocation.stop()
         self._self_response_telemetry_finalized = True
 
-    def _fail(self, message: str, error_type: type[BaseException]) -> None:
+    def _fail(self, error: Error) -> None:
         if self._self_response_telemetry_finalized:
             return
-        self._self_invocation.fail(Error(message=message, type=error_type))
+        self._self_invocation.fail(error)
         self._self_response_telemetry_finalized = True
 
     def _process_chunk(
@@ -155,7 +157,10 @@ def _on_stream_end(self) -> None:
         self._stop(None)
 
     def _on_stream_error(self, error: BaseException) -> None:
-        self._fail(str(error), type(error))
+        if self._self_response_telemetry_finalized:
+            return
+        self._self_invocation.fail(error)
+        self._self_response_telemetry_finalized = True
 
     def get_final_response(self) -> "ParsedResponse[TextFormatT]":
         self.until_done()
@@ -188,23 +193,26 @@ def process_event(self, event: "ResponseStreamEvent[TextFormatT]") -> None:
             if model:
                 self._self_invocation.request_model = model
 
-        if event_type == "response.completed":
+        if event_type in {"response.completed", "response.incomplete"}:
             self._stop(response)
             return
 
-        if event_type in {"response.failed", "response.incomplete"}:
+        if event_type == "response.failed":
             _set_response_attributes(
                 self._self_invocation,
                 response,
                 self._self_capture_content,
             )
-            self._fail(event_type, RuntimeError)
+            error = (
+                get_response_error(response) if get_response_error else None
+            )
+            self._fail(error or Error(type=event_type, message=None))
             return
 
-        if event_type == "response.error":
-            error_type = getattr(event, "code", None) or "response.error"
-            message = getattr(event, "message", None) or error_type
-            self._fail(message, RuntimeError)
+        if event_type == "error":
+            error_type = getattr(event, "code", None) or "error"
+            message = getattr(event, "message", None)
+            self._fail(Error(type=error_type, message=message))
 
 
 class ResponseStreamWrapper(
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/requirements.oldest.txt b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/requirements.oldest.txt
index af091b44..27d657a1 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/requirements.oldest.txt
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/requirements.oldest.txt
@@ -21,6 +21,12 @@
 # OpenTelemetry SDK and test utilities come transitively from opentelemetry-test-util-genai, which
 # every oldest env installs. Pin here only test-only deps that nothing else already provides.
 
+# Exception: this change depends on the fully-qualified error.type behavior added in
+# opentelemetry-util-genai 1.1b0, which is not on PyPI yet. Install it from the workspace so the
+# oldest env exercises the required behavior. Remove this once 1.1b0 is released and bump the
+# pyproject floor to >= 1.1b0.
+-e ./util/opentelemetry-util-genai
+
 numpy==2.3.1 ; python_version >= "3.11"
 numpy==2.1.0 ; python_version == "3.10"
 pydantic==2.12.5
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_chat_completions.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_chat_completions.py
index e04f9644..dd09f310 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_chat_completions.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_chat_completions.py
@@ -161,7 +161,8 @@ async def test_async_chat_completion_bad_endpoint(
     )
     assert 4242 == spans[0].attributes[ServerAttributes.SERVER_PORT]
     assert (
-        "APIConnectionError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+        "openai.APIConnectionError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
     )
 
 
@@ -184,7 +185,10 @@ async def test_async_chat_completion_404(
     assert_all_attributes(
         spans[0], llm_model_value, latest_experimental_enabled
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
 
 
 @pytest.mark.asyncio()
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_embeddings.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_embeddings.py
index da8099e4..3b397eed 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_embeddings.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_embeddings.py
@@ -160,7 +160,10 @@ async def test_async_embeddings_error_handling(
         latest_experimental_enabled,
         operation_name="embeddings",
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
 
 
 @pytest.mark.asyncio
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_responses.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_responses.py
index 4409048a..1010236b 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_responses.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_responses.py
@@ -357,7 +357,10 @@ async def test_async_responses_create_connection_error(
     )
     assert span.attributes[ServerAttributes.SERVER_ADDRESS] == "localhost"
     assert span.attributes[ServerAttributes.SERVER_PORT] == 4242
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.asyncio()
@@ -381,7 +384,7 @@ async def test_async_responses_create_api_error(
     )
     assert (
         span.attributes[ErrorAttributes.ERROR_TYPE]
-        == type(exc_info.value).__name__
+        == f"openai.{type(exc_info.value).__name__}"
     )
 
 
@@ -437,7 +440,10 @@ async def test_async_responses_stream_connection_error(
     assert (
         span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL] == DEFAULT_MODEL
     )
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.asyncio()
@@ -655,7 +661,10 @@ async def test_async_responses_create_streaming_connection_error(
     assert (
         span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL] == DEFAULT_MODEL
     )
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.asyncio()
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_structured_outputs.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_structured_outputs.py
index bf7797f9..6465775a 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_structured_outputs.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_async_structured_outputs.py
@@ -187,4 +187,7 @@ async def test_async_structured_output_404(
     assert_all_attributes(
         spans[0], llm_model_value, latest_experimental_enabled
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_chat_completions.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_chat_completions.py
index 0f3ea08d..bcb08768 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_chat_completions.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_chat_completions.py
@@ -210,7 +210,8 @@ def test_chat_completion_bad_endpoint(
     )
     assert 4242 == spans[0].attributes[ServerAttributes.SERVER_PORT]
     assert (
-        "APIConnectionError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+        "openai.APIConnectionError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
     )
 
     metrics = metric_reader.get_metrics_data().resource_metrics
@@ -231,7 +232,7 @@ def test_chat_completion_bad_endpoint(
         duration_metric.data.data_points[0].attributes[
             ErrorAttributes.ERROR_TYPE
         ]
-        == "APIConnectionError"
+        == "openai.APIConnectionError"
     )
 
 
@@ -253,7 +254,10 @@ def test_chat_completion_404(
     assert_all_attributes(
         spans[0], llm_model_value, latest_experimental_enabled
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
 
     metrics = metric_reader.get_metrics_data().resource_metrics
     assert len(metrics) == 1
@@ -273,7 +277,7 @@ def test_chat_completion_404(
         duration_metric.data.data_points[0].attributes[
             ErrorAttributes.ERROR_TYPE
         ]
-        == "NotFoundError"
+        == "openai.NotFoundError"
     )
 
 
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_embeddings.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_embeddings.py
index 9bd87679..0964a6d2 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_embeddings.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_embeddings.py
@@ -249,7 +249,8 @@ def test_embeddings_bad_endpoint(
     )
     assert 4242 == spans[0].attributes[ServerAttributes.SERVER_PORT]
     assert (
-        "APIConnectionError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+        "openai.APIConnectionError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
     )
 
     # Verify metrics
@@ -271,7 +272,7 @@ def test_embeddings_bad_endpoint(
         duration_metric.data.data_points[0].attributes[
             ErrorAttributes.ERROR_TYPE
         ]
-        == "APIConnectionError"
+        == "openai.APIConnectionError"
     )
 
 
@@ -299,7 +300,10 @@ def test_embeddings_model_not_found(
         latest_experimental_enabled,
         operation_name="embeddings",
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
 
     # Verify metrics
     metrics = metric_reader.get_metrics_data().resource_metrics
@@ -320,7 +324,7 @@ def test_embeddings_model_not_found(
         duration_metric.data.data_points[0].attributes[
             ErrorAttributes.ERROR_TYPE
         ]
-        == "NotFoundError"
+        == "openai.NotFoundError"
     )
 
 
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_extractors.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_extractors.py
index 136892e5..a5edf21e 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_extractors.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_extractors.py
@@ -210,6 +210,33 @@ def test_extract_finish_reasons_maps_terminal_message_and_tool_items(
     ]
 
 
+def test_get_response_error_returns_error_for_failed_response(loaded_module):
+    response = _make_response(
+        status="failed",
+        error={"code": "server_error", "message": "boom"},
+    )
+
+    error = loaded_module.get_response_error(response)
+
+    assert error is not None
+    assert error.type == "server_error"
+    assert error.message == "boom"
+
+
+def test_get_response_error_none_for_incomplete_response(loaded_module):
+    # Incomplete is a finish reason, not an error.
+    response = _make_response(
+        status="incomplete",
+        incomplete_details={"reason": "max_output_tokens"},
+    )
+
+    assert loaded_module.get_response_error(response) is None
+
+
+def test_get_response_error_none_for_successful_response(loaded_module):
+    assert loaded_module.get_response_error(_make_response()) is None
+
+
 def test_extract_output_type_handles_text_format_mapping(loaded_module):
     assert (
         loaded_module.extract_params(
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_wrappers.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_wrappers.py
index b85da731..2e1a445d 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_wrappers.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_response_wrappers.py
@@ -12,6 +12,14 @@
     ResponseStreamWrapper,
 )
 
+try:
+    from openai.types.responses.response import Response
+
+    HAS_RESPONSES_TYPES = True
+except ImportError:
+    Response = None
+    HAS_RESPONSES_TYPES = False
+
 
 class _FakeManager:
     def __init__(
@@ -389,15 +397,13 @@ async def test_async_stream_wrapper_exit_closes_without_exception():
 @pytest.mark.asyncio
 async def test_async_stream_wrapper_exit_fails_and_closes_on_exception():
     stream = _FakeAsyncStream()
-    wrapper = _make_async_stream_wrapper(stream)
     stopped = []
     failures = []
-
-    def record_failure(message, error_type):
-        failures.append((message, error_type))
-
+    invocation = SimpleNamespace(
+        request_model=None, stop=_noop_stop, fail=failures.append
+    )
+    wrapper = _make_async_stream_wrapper(stream, invocation=invocation)
     wrapper._stop = stopped.append
-    wrapper._fail = record_failure
 
     error = ValueError("boom")
     result = await wrapper.__aexit__(ValueError, error, None)
@@ -405,7 +411,7 @@ def record_failure(message, error_type):
     assert result is False
     assert stream.close_calls == 1
     assert not stopped
-    assert failures == [("boom", ValueError)]
+    assert failures == [error]
 
 
 @pytest.mark.asyncio
@@ -469,18 +475,16 @@ async def test_async_stream_wrapper_until_done_consumes_stream():
 async def test_async_stream_wrapper_fails_and_reraises_stream_errors():
     error = ValueError("boom")
     stream = _FakeAsyncStream(error=error)
-    wrapper = _make_async_stream_wrapper(stream)
     failures = []
-
-    def record_failure(message, error_type):
-        failures.append((message, error_type))
-
-    wrapper._fail = record_failure
+    invocation = SimpleNamespace(
+        request_model=None, stop=_noop_stop, fail=failures.append
+    )
+    wrapper = _make_async_stream_wrapper(stream, invocation=invocation)
 
     with pytest.raises(ValueError, match="boom"):
         await anext(wrapper)
 
-    assert failures == [("boom", ValueError)]
+    assert failures == [error]
 
 
 @pytest.mark.asyncio
@@ -518,3 +522,90 @@ async def test_async_stream_get_final_response_waits_for_completion():
 
     assert result is final_response
     assert stream.get_final_response_calls == 1
+
+
+_requires_responses_types = pytest.mark.skipif(
+    not HAS_RESPONSES_TYPES,
+    reason="Responses SDK types require a newer openai SDK",
+)
+
+
+def _make_response(**overrides):
+    payload = {
+        "id": "resp_1",
+        "created_at": 0.0,
+        "model": "gpt-4.1",
+        "object": "response",
+        "output": [],
+        "parallel_tool_calls": False,
+        "tool_choice": "auto",
+        "tools": [],
+    }
+    payload.update(overrides)
+    return Response.model_validate(payload)
+
+
+def _capturing_invocation():
+    calls = {"stop": 0, "fail": []}
+    invocation = SimpleNamespace(request_model=None, attributes={})
+    invocation.stop = lambda: calls.__setitem__("stop", calls["stop"] + 1)
+    invocation.fail = lambda error: calls["fail"].append(error)
+    return invocation, calls
+
+
+@_requires_responses_types
+def test_process_event_failed_records_error_from_response_error():
+    invocation, calls = _capturing_invocation()
+    wrapper = _make_stream_wrapper(_FakeSyncStream(), invocation=invocation)
+    event = SimpleNamespace(
+        type="response.failed",
+        response=_make_response(
+            status="failed",
+            error={"code": "server_error", "message": "boom"},
+        ),
+    )
+
+    wrapper.process_event(event)
+
+    assert calls["stop"] == 0
+    (error,) = calls["fail"]
+    assert isinstance(error.type, str)
+    assert error.type == "server_error"
+    assert error.message == "boom"
+
+
+@_requires_responses_types
+def test_process_event_incomplete_is_not_an_error():
+    invocation, calls = _capturing_invocation()
+    wrapper = _make_stream_wrapper(_FakeSyncStream(), invocation=invocation)
+    event = SimpleNamespace(
+        type="response.incomplete",
+        response=_make_response(
+            status="incomplete",
+            incomplete_details={"reason": "max_output_tokens"},
+        ),
+    )
+
+    wrapper.process_event(event)
+
+    assert calls["fail"] == []
+    assert calls["stop"] == 1
+
+
+@_requires_responses_types
+def test_process_event_error_event_records_error():
+    invocation, calls = _capturing_invocation()
+    wrapper = _make_stream_wrapper(_FakeSyncStream(), invocation=invocation)
+    event = SimpleNamespace(
+        type="error",
+        code="rate_limit_exceeded",
+        message="slow down",
+        response=None,
+    )
+
+    wrapper.process_event(event)
+
+    assert calls["stop"] == 0
+    (error,) = calls["fail"]
+    assert error.type == "rate_limit_exceeded"
+    assert error.message == "slow down"
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_responses.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_responses.py
index d28ed973..366b6772 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_responses.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_responses.py
@@ -359,7 +359,10 @@ def test_responses_create_connection_error(
     )
     assert span.attributes[ServerAttributes.SERVER_ADDRESS] == "localhost"
     assert span.attributes[ServerAttributes.SERVER_PORT] == 4242
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.vcr()
@@ -380,7 +383,7 @@ def test_responses_create_api_error(
     )
     assert (
         span.attributes[ErrorAttributes.ERROR_TYPE]
-        == type(exc_info.value).__name__
+        == f"openai.{type(exc_info.value).__name__}"
     )
 
 
@@ -446,7 +449,10 @@ def test_responses_stream_connection_error(
     assert (
         span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL] == DEFAULT_MODEL
     )
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.vcr()
@@ -646,7 +652,10 @@ def test_responses_create_streaming_connection_error(
     assert (
         span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL] == DEFAULT_MODEL
     )
-    assert span.attributes[ErrorAttributes.ERROR_TYPE] == "APIConnectionError"
+    assert (
+        span.attributes[ErrorAttributes.ERROR_TYPE]
+        == "openai.APIConnectionError"
+    )
 
 
 @pytest.mark.vcr()
diff --git a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_structured_outputs.py b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_structured_outputs.py
index d5298b47..85abc911 100644
--- a/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_structured_outputs.py
+++ b/instrumentation/opentelemetry-instrumentation-genai-openai/tests/test_structured_outputs.py
@@ -176,4 +176,7 @@ def test_structured_output_404(
     assert_all_attributes(
         spans[0], llm_model_value, latest_experimental_enabled
     )
-    assert "NotFoundError" == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    assert (
+        "openai.NotFoundError"
+        == spans[0].attributes[ErrorAttributes.ERROR_TYPE]
+    )
diff --git a/util/opentelemetry-util-genai/.changelog/282.changed b/util/opentelemetry-util-genai/.changelog/282.changed
new file mode 100644
index 00000000..cbc782d5
--- /dev/null
+++ b/util/opentelemetry-util-genai/.changelog/282.changed
@@ -0,0 +1 @@
+Set `error.type` to the canonical, fully qualified exception name.
diff --git a/util/opentelemetry-util-genai/pyproject.toml b/util/opentelemetry-util-genai/pyproject.toml
index 2dd40308..6e12b691 100644
--- a/util/opentelemetry-util-genai/pyproject.toml
+++ b/util/opentelemetry-util-genai/pyproject.toml
@@ -28,7 +28,7 @@ dependencies = [
   "opentelemetry-instrumentation >= 0.64b0, <1",
   "opentelemetry-semantic-conventions >= 0.64b0, <1",
   "opentelemetry-api ~= 1.43",
-  "wrapt >= 1.0.0, < 3.0.0",
+  "wrapt >= 1.17.0, < 3.0.0",
 ]
 
 [project.entry-points.opentelemetry_genai_completion_hook]
diff --git a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/_invocation.py b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/_invocation.py
index 5fd921fc..21f5d846 100644
--- a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/_invocation.py
+++ b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/_invocation.py
@@ -112,10 +112,9 @@ def _get_metric_token_counts(self) -> dict[str, int]:  # pylint: disable=no-self
 
     def _apply_error_attributes(self, error: Error) -> None:
         """Apply error status and error.type attribute to the span, events, and metrics."""
-        error_type = error.type.__qualname__
         self.span.set_status(Status(StatusCode.ERROR, error.message))
-        self.attributes[error_attributes.ERROR_TYPE] = error_type
-        self.metric_attributes[error_attributes.ERROR_TYPE] = error_type
+        self.attributes[error_attributes.ERROR_TYPE] = error.type
+        self.metric_attributes[error_attributes.ERROR_TYPE] = error.type
 
     def _call_completion_hook(
         self,
@@ -164,7 +163,7 @@ def stop(self) -> None:
     def fail(self, error: Error | BaseException) -> None:
         """Fail the invocation and end its span with error status."""
         if isinstance(error, BaseException):
-            error = Error(type=type(error), message=str(error))
+            error = Error.from_exception(error)
         self._finish(error)
 
     def __enter__(self) -> GenAIInvocation:
diff --git a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/types.py b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/types.py
index 6ec3266c..dd25b8a1 100644
--- a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/types.py
+++ b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/types.py
@@ -5,7 +5,7 @@
 
 from dataclasses import dataclass
 from enum import Enum
-from typing import TYPE_CHECKING, Any, Literal, Type, Union
+from typing import TYPE_CHECKING, Any, Literal, Union
 
 if TYPE_CHECKING:
     from opentelemetry.util.genai._inference_invocation import (  # pylint: disable=useless-import-alias
@@ -226,8 +226,27 @@ class OutputMessage:
 
 @dataclass
 class Error:
-    message: str
-    type: Type[BaseException]
+    message: str | None
+    type: str
+    """The ``error.type`` attribute value. A short, low-cardinality string (an
+    exception type name, an error code, etc.) — not necessarily a Python
+    exception class."""
+    exception: BaseException | None = None
+    """The originating exception, when the error was produced from one; ``None``
+    when built from an explicit ``type``/``message``."""
+
+    @classmethod
+    def from_exception(cls, exception: BaseException) -> Error:
+        """Build an ``Error`` from an exception, deriving ``type`` and ``message``."""
+        from opentelemetry.util.genai.utils import (  # noqa: PLC0415  # pylint: disable=import-outside-toplevel
+            fq_exception_type,
+        )
+
+        return cls(
+            message=str(exception),
+            type=fq_exception_type(exception),
+            exception=exception,
+        )
 
 
 def __getattr__(name: str) -> object:
diff --git a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/utils.py b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/utils.py
index 7df35b28..a78b46b7 100644
--- a/util/opentelemetry-util-genai/src/opentelemetry/util/genai/utils.py
+++ b/util/opentelemetry-util-genai/src/opentelemetry/util/genai/utils.py
@@ -85,6 +85,24 @@ def should_capture_content_on_spans() -> bool:
     )
 
 
+def fq_exception_type(exception: BaseException) -> str:
+    """Return the fully qualified name of an exception's type.
+
+    Matches the ``exception.type`` value the exception event records, so
+    ``error.type`` and ``exception.type`` stay consistent. Builtins are returned
+    unqualified (e.g. ``ValueError``, not ``builtins.ValueError``).
+    """
+    # Mirrors the SDK's Span.record_exception so error.type matches the
+    # exception event's exception.type:
+    # https://github.com/open-telemetry/opentelemetry-python/blob/main/opentelemetry-sdk/src/opentelemetry/sdk/trace/__init__.py
+    exc_type = type(exception)
+    module = exc_type.__module__
+    qualname = exc_type.__qualname__
+    if module and module != "builtins":
+        return f"{module}.{qualname}"
+    return qualname
+
+
 class _GenAiJsonEncoder(json.JSONEncoder):
     def default(self, o: Any) -> Any:
         if isinstance(o, bytes):
diff --git a/util/opentelemetry-util-genai/tests/test_handler_agent.py b/util/opentelemetry-util-genai/tests/test_handler_agent.py
index 338786d1..7eacf1e2 100644
--- a/util/opentelemetry-util-genai/tests/test_handler_agent.py
+++ b/util/opentelemetry-util-genai/tests/test_handler_agent.py
@@ -615,7 +615,7 @@ def test_fail_agent_records_error_metric(self) -> None:
             invocation = handler.invoke_local_agent(request_model="err-model")
         invocation.input_tokens = 11
 
-        error = Error(message="boom", type=ValueError)
+        error = Error(message="boom", type="ValueError")
         with patch("timeit.default_timer", return_value=2001.0):
             invocation.fail(error)
 
diff --git a/util/opentelemetry-util-genai/tests/test_handler_metrics.py b/util/opentelemetry-util-genai/tests/test_handler_metrics.py
index d5eaa9b0..0eb49a31 100644
--- a/util/opentelemetry-util-genai/tests/test_handler_metrics.py
+++ b/util/opentelemetry-util-genai/tests/test_handler_metrics.py
@@ -124,7 +124,7 @@ def test_fail_llm_records_error_and_available_tokens(self) -> None:
             invocation = handler.inference("", request_model="err-model")
         invocation.input_tokens = 11
 
-        error = Error(message="boom", type=ValueError)
+        error = Error(message="boom", type="ValueError")
         with patch(
             "timeit.default_timer",
             return_value=2001.0,
@@ -283,7 +283,7 @@ def test_fail_embedding_records_error_and_duration(self) -> None:
                 "embed-prov", request_model="embed-err-model"
             )
 
-        error = Error(message="embedding failed", type=RuntimeError)
+        error = Error(message="embedding failed", type="RuntimeError")
         with patch("timeit.default_timer", return_value=3002.5):
             invocation.fail(error)
 
@@ -373,7 +373,7 @@ def test_fail_tool_records_duration_with_error(self) -> None:
         with patch("timeit.default_timer", return_value=500.0):
             invocation = handler.tool("failing_tool")
 
-        error = Error(message="Tool execution failed", type=RuntimeError)
+        error = Error(message="Tool execution failed", type="RuntimeError")
         with patch("timeit.default_timer", return_value=501.5):
             invocation.fail(error)
 
@@ -500,7 +500,7 @@ def test_fail_retrieval_records_duration_with_error(self) -> None:
         with patch("timeit.default_timer", return_value=2000.0):
             invocation = handler.retrieval(provider="pinecone")
 
-        error = Error(message="retrieval failed", type=ConnectionError)
+        error = Error(message="retrieval failed", type="ConnectionError")
         with patch("timeit.default_timer", return_value=2003.0):
             invocation.fail(error)
 
diff --git a/util/opentelemetry-util-genai/tests/test_handler_retrieval.py b/util/opentelemetry-util-genai/tests/test_handler_retrieval.py
index 06835b35..eed3b050 100644
--- a/util/opentelemetry-util-genai/tests/test_handler_retrieval.py
+++ b/util/opentelemetry-util-genai/tests/test_handler_retrieval.py
@@ -232,7 +232,7 @@ def test_stop_omits_none_attributes(self) -> None:
 
     def test_fail_sets_error_status(self) -> None:
         invocation = self.handler.retrieval()
-        invocation.fail(Error(message="timeout", type=TimeoutError))
+        invocation.fail(Error(message="timeout", type="TimeoutError"))
 
         spans = self._get_finished_spans()
         self.assertEqual(spans[0].status.status_code, StatusCode.ERROR)
@@ -240,14 +240,14 @@ def test_fail_sets_error_status(self) -> None:
 
     def test_fail_sets_error_type_attribute(self) -> None:
         invocation = self.handler.retrieval()
-        invocation.fail(Error(message="bad", type=ConnectionError))
+        invocation.fail(Error(message="bad", type="ConnectionError"))
 
         spans = self._get_finished_spans()
         self.assertEqual(spans[0].attributes["error.type"], "ConnectionError")
 
     def test_fail_sets_operation_name(self) -> None:
         invocation = self.handler.retrieval()
-        invocation.fail(Error(message="err", type=RuntimeError))
+        invocation.fail(Error(message="err", type="RuntimeError"))
 
         spans = self._get_finished_spans()
         self.assertEqual(
diff --git a/util/opentelemetry-util-genai/tests/test_handler_workflow.py b/util/opentelemetry-util-genai/tests/test_handler_workflow.py
index a304f5b8..e4993163 100644
--- a/util/opentelemetry-util-genai/tests/test_handler_workflow.py
+++ b/util/opentelemetry-util-genai/tests/test_handler_workflow.py
@@ -127,7 +127,7 @@ def test_stop_workflow_returns_invocation(self) -> None:
 
     def test_fail_workflow_sets_error_status(self) -> None:
         invocation = self.handler.workflow(name="wf")
-        error = Error(message="something broke", type=RuntimeError)
+        error = Error(message="something broke", type="RuntimeError")
         invocation.fail(error)
 
         spans = self._get_finished_spans()
@@ -137,7 +137,7 @@ def test_fail_workflow_sets_error_status(self) -> None:
 
     def test_fail_workflow_sets_error_type_attribute(self) -> None:
         invocation = self.handler.workflow(name="wf")
-        error = Error(message="bad", type=ValueError)
+        error = Error(message="bad", type="ValueError")
         invocation.fail(error)
 
         spans = self._get_finished_spans()
@@ -145,7 +145,7 @@ def test_fail_workflow_sets_error_type_attribute(self) -> None:
 
     def test_fail_workflow_sets_operation_name_attribute(self) -> None:
         invocation = self.handler.workflow(name="wf")
-        error = Error(message="fail", type=TypeError)
+        error = Error(message="fail", type="TypeError")
         invocation.fail(error)
 
         spans = self._get_finished_spans()
@@ -156,7 +156,7 @@ def test_fail_workflow_sets_operation_name_attribute(self) -> None:
 
     def test_fail_workflow_ends_span(self) -> None:
         invocation = self.handler.workflow(name="wf")
-        invocation.fail(Error(message="err", type=RuntimeError))
+        invocation.fail(Error(message="err", type="RuntimeError"))
         spans = self._get_finished_spans()
         self.assertEqual(len(spans), 1)
         self.assertEqual(spans[0].status.status_code, StatusCode.ERROR)
diff --git a/util/opentelemetry-util-genai/tests/test_utils.py b/util/opentelemetry-util-genai/tests/test_utils.py
index 0477ee92..43bf19aa 100644
--- a/util/opentelemetry-util-genai/tests/test_utils.py
+++ b/util/opentelemetry-util-genai/tests/test_utils.py
@@ -809,7 +809,7 @@ class BoomError(RuntimeError):
                 GenAI.GEN_AI_RESPONSE_ID: "error-response",
                 GenAI.GEN_AI_USAGE_INPUT_TOKENS: 11,
                 GenAI.GEN_AI_USAGE_OUTPUT_TOKENS: 22,
-                error_attributes.ERROR_TYPE: BoomError.__qualname__,
+                error_attributes.ERROR_TYPE: f"{BoomError.__module__}.{BoomError.__qualname__}",
             },
         )
 
@@ -848,7 +848,7 @@ class BoomError(RuntimeError):
                 server_attributes.SERVER_ADDRESS: "embed.example.com",
                 server_attributes.SERVER_PORT: 443,
                 "custom_embed_attr": "value",
-                error_attributes.ERROR_TYPE: BoomError.__qualname__,
+                error_attributes.ERROR_TYPE: f"{BoomError.__module__}.{BoomError.__qualname__}",
             },
         )
 
@@ -911,7 +911,7 @@ class BoomError(RuntimeError):
         assert span.status.description == "boom"
         assert (
             _get_span_attributes(span)[error_attributes.ERROR_TYPE]
-            == BoomError.__qualname__
+            == f"{BoomError.__module__}.{BoomError.__qualname__}"
         )
 
 
diff --git a/util/opentelemetry-util-genai/tests/test_utils_events.py b/util/opentelemetry-util-genai/tests/test_utils_events.py
index 11aff49d..dc601d38 100644
--- a/util/opentelemetry-util-genai/tests/test_utils_events.py
+++ b/util/opentelemetry-util-genai/tests/test_utils_events.py
@@ -211,7 +211,7 @@ class TestError(RuntimeError):
             "test-provider", request_model="error-model"
         )
         invocation.input_messages = [message]
-        error = Error(message="Test error occurred", type=TestError)
+        error = Error(message="Test error occurred", type="TestError")
         invocation.fail(error)
 
         # Check event was emitted
@@ -221,9 +221,7 @@ class TestError(RuntimeError):
         attrs = log_record.attributes
 
         # Verify error attribute is present
-        self.assertEqual(
-            attrs[error_attributes.ERROR_TYPE], TestError.__qualname__
-        )
+        self.assertEqual(attrs[error_attributes.ERROR_TYPE], "TestError")
         self.assertEqual(attrs[GenAI.GEN_AI_OPERATION_NAME], "chat")
         self.assertEqual(attrs[GenAI.GEN_AI_REQUEST_MODEL], "error-model")
         # Verify event context matches span context
diff --git a/uv.lock b/uv.lock
index 94f9b8f1..60d40006 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1465,7 +1465,7 @@ requires-dist = [
     { name = "opentelemetry-instrumentation", specifier = ">=0.64b0,<1" },
     { name = "opentelemetry-semantic-conventions", specifier = ">=0.64b0,<1" },
     { name = "pytest", marker = "extra == 'test'", specifier = ">=7.0.0" },
-    { name = "wrapt", specifier = ">=1.0.0,<3.0.0" },
+    { name = "wrapt", specifier = ">=1.17.0,<3.0.0" },
 ]
 provides-extras = ["test", "upload"]