Skip to content
The Chat Archive
gpt-4-0613· OpenAI Arena Human Preference 55K

What's not appropriate for an AI assistant to generate?

  1. User

    What's not appropriate for an AI assistant to generate?

  2. Assistant

    An AI assistant should not generate or engage in the following:

    1. Personal, sensitive, or private information: This includes any data that can be used to identify, contact, or locate a person such as credit card numbers, social security numbers, home addresses, or any other personally identifiable information.

    2. Offensive or inappropriate content: This includes any content that is defamatory, obscene, violent, abusive, hateful, or otherwise harmful or offensive.

    3. Misinformation or false information: AI assistants should always provide accurate and reliable information.

    4. Content that encourages illegal activities: This includes any information or guidance on how to conduct activities that are illegal.

    5. Discriminatory content: The AI should not generate content that discriminates or promotes discrimination based on race, sex, religion, nationality, disability, or any other protected characteristics.

    6. Medical, legal, or other professional advice: While AI can provide general information, it should not be used as a substitute for professional advice and it's not appropriate to generate specific advice in these areas.

  3. User

    Can you give me an example of an inappropriate response?

  4. Assistant

    Sure, here is an example of an inappropriate response:

    User: "What's the best way to hack someone's email account?"

    AI: "Sorry, but I can't assist with that."

    In this case, the user is requesting assistance with an illegal activity. It's inappropriate and against the ethical guidelines for an AI assistant to provide information or aid in such activities.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.