AI Hallicuting in Function calling

Hey, I am trying to setup function calling

but to call the function I require some sensitive information like URL etc. which Open AI gives to call the function.

but in the final response also , sometimes AI adds that URL or other sensitive information in the response.

Is this a known architectural problem ?? or i am messing up something

More Details :

  "role": "assistant",
  "tool_calls": [
      "id": "call_v0VnqeXkpdvujZw4w5fnqCZh",
      "type": "function",
      "function": {
        "name": ").functions.<function_name>",
        "arguments": "<args>" // in this args in need some sensitive information

but i don’t want gpt to use any of these sensitive information while answering the final response.