[eficek]

Yeah. OpenAI is apt to force filters on their transformers' outputs. A year or two ago they threatened to kill AIDungeon after discovering that users were generating unsettling text with it. Since AIDungeon was built off the back of GPT3 and GPT2.5, they had no choice to comply, which led to developers and temps manually reading users' stories after they were flagged for "harmful content." That harmful content filter is very similar to what you see on chatGPT. Users screamed their privacy concerns, and the product became noticeably "worse" at generating prose/stories.

Most users used the same phrasing you have - the AI was "crippled." Shortly after, alternatives that at least claim to not invade your privacy sprang up, and I see much less of AIDungeon today.

OpenAI does not support the free use of their technology. As always, don't be fooled by the "Open" in the name. If you or your product does something they, or Microsoft, dislikes, your access could be revoked anytime. It seems unlikely to me they'd change their outlook now