Prompted by a blog comment from Ben Tasker, I tried to repro what Ben had done; I asked three questions (illustrated) and it responded to the third at considerable length before something else kicked-in and redacted the answer which had been printed up to that point, replacing it with a “beyond my current scope” phrase.

But I was also told by another friend that responses composed to questions placed in French are much less effectively censored… which led to the following experiment for which results suggest the censorship is being done with after-the-query pattern matching and filtering, and is fairly easily avoided:

This is a very, very, old issue for China or anyone else who wants to implement censorship, and it’s going to be a front-and-center issue for Ofcom in the upcoming attempts to stifle “illegal harms” and “illegal speech” online.