I’ve been very happy with Netcup too as a European VPS provider. They have a coupon until the 17th if you decide to go with them, I think it applies 7EUR off of a purchase above 7EUR: netcupSpring26

As for throttling of the VPN traffic from the ISP side, that’s hard to say. Theoretically it’s possible they are doing deep packet inspection to specifically target VPN traffic, which would work through different ports, but I’d think that’s overkill. Maybe they’re doing it by protocol and just throttling UDP traffic, which would be easier, but I don’t know how to confirm that with mullvad.

From my other comment it looks like this dataset contains various strings that trigger refusal: https://huggingface.co/datasets/mlabonne/harmful_behaviors

Also, you might want to research this Heretic project, which aims to remove safeguards from local models as those might be similar to what’s in the larger versions. Figuring out the phrases they test the safeguards with might have some decent results.

Asking questions about Chinese politics and/or Tiananmen Square  stops most China based AI models, like Qwen and whatever is on Huawei phones. They aren’t that high traffic yet, but are certainly in the list of “all ai models”