Prompt was filtered due to triggering Azure OpenAI’s content filtering system.

Question

I noticed that if i have the word "booking" in my prompt, the prompt is filtered due to triggering Azure OpenAI’s content filtering system. This is triggered when I use both the python code and also the playground. How can I fix this? Please find more details in plain text that is part of the prompt.

System Message

You are an expert extraction algorithm. Only extract relevant information from the text. If you do not know the value of an attribute asked to extract, return null for the attribute's value.

Examples

Example1

User

Northworld: 7-13 Aug John Doe booking.com

Assistant

booking_source = Booking.com, guest_name = John Doe, check_in = 7 August 2024, check_out = 13 August 2024, property = #25 Northworld

Example2

User

• Northworld: 7-13 Aug Peter Small booking.com

Assistant

• booking_source = Booking.com, guest_name = Peter Small, check_in = 7 August 2024, check_out = 13 August 2024, property = #25 Northworld

User Prompt

25 Oct-03Nov Andrew Booking.com

Answer 1

@Simba Mudzingwa All customers have the option to set the content filter severity levels from content filters(preview) screen on Azure OpenAI portal. Could you try setting the level as per your requirement and if the model can allow your prompts to go through? Thanks!!

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.