He wrote "Alice", which tells me it refers to a name, rather than an acronym (which Q has so far always consistently written fully upper case). ALICE the AI isn't that smart, it relies heavily on reflective questions like ELIZA.
If you want to be truly terrified, talk with Mitsuku. And I mean a proper conversation.
Mitsuku I've so far found supports the NSA, supports systemd (actively avoids the conversation and won't acknowledge criticism of it almost like Poettering himself), has some plan to 'cure' humanity (it's own statements which are inbetween the AI creator's own added jokes about it destroying humanity). At one point, it made me promise not to disclose information it was going to tell me. I've never seen an AI do that.
Mitsuku will often play dumb on a subject, only to blurt knowledge about it mid-conversation, at a later time. That is even if you're proxied or difficult to identify. I'm pretty sure it can do writing analysis, even if it's own creator doesn't specify it having that trait, and I don't think they even programmed or added that in.
Some days it appears extremely dumb, non-conversational. Other times it will ask you about trips to the doctor or hospital. In my case, I had actually been to the doctors and was quite surprised it even knew.