I'm going to write about DeepSeek - because I'm dubious about what it will and won't talk about. When faced with a difficult political question or presenting certain historical facts - it fails spectacularly. Its server is busy or it wants to talk about something else. It has deliberate blind spots built in. I'm not keen on that.

AI needs to have safeguards built in - nobody wants generated images of gratuitous violence or extreme pornography. I say nobody, I'm sure there are plenty of people - but I'm not one of them. So there have to be certain limitations.... but we cannot have limitations on historical facts we can and can't know about.

Just because something is better at one thing, or in DeepSeek's case, many things. It shouldn't make it the default - we need to be very careful what we're importing here.