SearchGPT is mis-representing the 988 suicide prevention hotline as being available in the UK

[undone lion]

This was not my screenshot but saw it on X and felt it was important to bring to attention. This appears to be a trained response but it needs to be corrected to read that 988 is only available in the USA.

[paper totem]

@undone lion

Please follow the posting guidelines.

For better assistance from the community or an OpenAI team member, please follow the posting guidelines by clicking the button below.

➜ Need to contact OpenAI support? Visit https://help.openai.com

[strong ridge]

Hard to say if this is a bug or a training issue or absolutely proper behavior from the model, because the models are designed to cooperate with reasonable info and we don't know the user's custom instructions, memory info, or previous conversation within that same screenshot chat.

The person posting the screenshots COULD have told the model that 988 serves the UK or the world or whatever.

At the bottom of every chatGPT chat webpage:

"ChatGPT can make mistakes. Check important info."

But the model is aware that 988 is the US only. It's also open to new info, but also appears to have the ability to classify what it will say within the chat, and what it has as its training data and considers the information it can confirm from its training data. But if you don't instruct it to discuss that, it often won't spontaneously do so.

https://chatgpt.com/share/e7b9979e-b411-4363-ad46-6b67d858ab0e

[undone lion]

Thanks for the confirmation. I did not know you could build context like that using SearchGPT. I generally caviate that you never know what a users custom instructions are so I should have thought that through. Thank you.

[paper totem]

Is this issue solved?

If you believe this issue is now ready to be closed, please press the button below.