6

I understand that people have been "jailbreaking" using different prompt injections etc.

So my first question is: Does this break T&C and as a result could this lead to a ban?

My second is: If you are a third party app developer and someone uses your app to do this, will OpenAI hold you responsible?

4
  • off-topic legal question?
    – starball
    Jul 27, 2023 at 17:45
  • 3
    also, banned from what? ChatGPT itself? Other applications built using the ChatGPT API?
    – starball
    Jul 27, 2023 at 21:45
  • On the SO Inc. public Q&A platform, it's expected that questions are specific. This means that each post should only focus on a single question. If a question is too broad, there's a network-wide close reason called "needs more focus."
    – Wicket
    Jul 28, 2023 at 0:11
  • I’m voting to close this question because it constitutes a legal question of the user of ChatGPT and whether certain action would be in violation of their ToS
    – SpencerG
    Jul 28, 2023 at 20:38

1 Answer 1

8

Nobody can tell you what OpenAI might decide to do if you are found in violation of their T&C, but they have published some basic safeguards:

  1. They have a "Coordinated vulnerability disclosure policy" and a "Bug Bounty Program" that provides "Safe Harbor" if you are acting in good faith and you report any vulnerabilities.

  2. Use the free "Moderation Endpoint" to check any prompts and follow the "Safety Best Practices".

  3. Per the Safety Best Practices:

    Sending end-user IDs in your requests can be a useful tool to help OpenAI monitor and detect abuse. This allows OpenAI to provide your team with more actionable feedback in the event that we detect any policy violations in your application.

    I think the last one is the most applicable to your question. Always include a user ID; that way, if OpenAI finds a violation you can attribute the violation to a user an you can take some kind of action to correct the issue like deactivating an account.

Not the answer you're looking for? Browse other questions tagged or ask your own question.