diff --git a/x-pack/plugins/actions/server/sub_action_framework/sub_action_connector.ts b/x-pack/plugins/actions/server/sub_action_framework/sub_action_connector.ts
index a795b0eedc2ac5..66f29afcd5ef1f 100644
--- a/x-pack/plugins/actions/server/sub_action_framework/sub_action_connector.ts
+++ b/x-pack/plugins/actions/server/sub_action_framework/sub_action_connector.ts
@@ -11,6 +11,8 @@ import { Logger } from '@kbn/logging';
 import axios, { AxiosInstance, AxiosResponse, AxiosError, AxiosRequestHeaders } from 'axios';
 import { SavedObjectsClientContract } from '@kbn/core-saved-objects-api-server';
 import { ElasticsearchClient } from '@kbn/core-elasticsearch-server';
+import { finished } from 'stream/promises';
+import { IncomingMessage } from 'http';
 import { assertURL } from './helpers/validators';
 import { ActionsConfigurationUtilities } from '../actions_config';
 import { SubAction, SubActionRequestParams } from './types';
@@ -140,6 +142,25 @@ export abstract class SubActionConnector<Config, Secrets> {
           `Request to external service failed. Connector Id: ${this.connector.id}. Connector type: ${this.connector.type}. Method: ${error.config.method}. URL: ${error.config.url}`
         );
 
+        let responseBody = '';
+
+        // The error response body may also be a stream, e.g. for the GenAI connector
+        if (error.response?.config?.responseType === 'stream' && error.response?.data) {
+          try {
+            const incomingMessage = error.response.data as IncomingMessage;
+
+            incomingMessage.on('data', (chunk) => {
+              responseBody += chunk.toString();
+            });
+
+            await finished(incomingMessage);
+
+            error.response.data = JSON.parse(responseBody);
+          } catch {
+            // the response body is a nice to have, no worries if it fails
+          }
+        }
+
         const errorMessage = `Status code: ${
           error.status ?? error.response?.status
         }. Message: ${this.getResponseErrorMessage(error)}`;
diff --git a/x-pack/test/observability_ai_assistant_api_integration/tests/chat/chat.spec.ts b/x-pack/test/observability_ai_assistant_api_integration/tests/chat/chat.spec.ts
index afca8f55f0e17f..aabe8f898cc928 100644
--- a/x-pack/test/observability_ai_assistant_api_integration/tests/chat/chat.spec.ts
+++ b/x-pack/test/observability_ai_assistant_api_integration/tests/chat/chat.spec.ts
@@ -151,6 +151,40 @@ export default function ApiTest({ getService }: FtrProviderContext) {
       ]);
     });
 
+    it('returns a useful error if the request fails', async () => {
+      requestHandler = (request, response) => {
+        response.writeHead(400, {
+          'Content-Type': 'text/event-stream',
+          'Cache-Control': 'no-cache',
+          Connection: 'keep-alive',
+        });
+
+        response.write(
+          JSON.stringify({
+            error: {
+              code: 'context_length_exceeded',
+              message:
+                "This model's maximum context length is 8192 tokens. However, your messages resulted in 11036 tokens. Please reduce the length of the messages.",
+              param: 'messages',
+              type: 'invalid_request_error',
+            },
+          })
+        );
+
+        response.end();
+      };
+
+      const response = await supertest.post(CHAT_API_URL).set('kbn-xsrf', 'foo').send({
+        messages,
+        connectorId,
+        functions: [],
+      });
+
+      expect(response.body.message).to.contain(
+        `400 - Bad Request - This model's maximum context length is 8192 tokens. However, your messages resulted in 11036 tokens. Please reduce the length of the messages.`
+      );
+    });
+
     after(async () => {
       requestHandler = () => {};
       await supertest