What a GPT ban does do is give moderators an extremely subjective tool for removing answers. The GPT detectors are unreliable, so that leaves the moderators doing a gut check on whether this particular case is a false positive.
Instead, why don't they just use the GPT detectors to detect answers that might be wrong and then moderate based on existing policies?
Now spread that one bot over hundreds of users and you have the same end result. That's why the communities of different SE sites all ended up with a similar policy.
I'm not suggesting either problem is easy, but taking the "ban AI" approach is harmful for several reasons. Detecting AI without false positives is extremely hard and puts too much decision-making power in the gut reactions of moderators. Additionally, banning it has the negative side effect of making the tool unavailable to people who might legitimately benefit from it when producing good content, such as ESL speakers using it to polish their English.
We need to moderate based on the effect of the user's content on the community, not the technology used to produce the content. If you're dealing with a bot posting 10k questions and answers per hour, it wouldn't matter if that bot were using GPT or just re-posting content scraped from the web—the abuse isn't in the use of AI, it's in the spam. So make a rule against spam and auto-ban people who do it.