They're using it to a. See what exactly are red flags to us such that we notice them as non-human, b. Testing to see when they successfully slip past us.
These are the obvious ones; I suspect there are many that we are now not noticing.
We're providing free data to them whenever we do or do not flag the comments, unfortunately.
Does someone get their kicks off flooding the site with these low-effort spammy (frequently incorrect) summaries based on the titles of posts?
If I wanted to be subjected to AI slop I'd just paste the title into ChatGPT myself.