This content might violate our prompt policy

[kindred cypress]

I'm working on a transcript from my podcast to create a short version of the text and constantly get this error message on content policy.

[inland eagle]

Yeah, GPT is overly sensitive

There's certain topics that just can't be discussed whatsoever

[keen lintel]

I only triggered this once, and it was just a “might” violate, so I am not really sure what people are doing to trigger it

I’m always a little suspicious when someone gets this

Maybe it’s sensitive about specific topics I don’t use though

[weary oracle]

I just an email about this as well. Tracing it through the prompt that was flagged was specific for news summarisation. The thing is, the output is merged on the client side with the adequate disclaimer as per the content policies so I don't know what is expected here as per the terms...

[limber olive]

seems like a "working as designed" issue to me. context is blurry and it's failing conservatively towards safety.

Not meant to be spam - just something to actually consider.

[sick river]

That you investigate and verify that your content hasn't violated terms and, if it has, take steps to prevent it.

[weary oracle]

There's no culminate effect with these issues? If its just an alert to check and investigate thats fine. If its a warning then there's a subtle undercurrent whereby developers can loose access for innocent meaning and over conservative "fail safe" design. Fail safe isn't the only failure mode after all.