They don't now, but enough "high value to the bots" pages turning on JS or complicated redirects will simply result in the bot authors adding JS execution or redirect following so they can continue "botting" the sites they want to scrape.
It's a hole with no bottom. Each one-up on the anti-bot side will eventually be handled on the bot side.
And there are just not enough sites using Anubis for the people and companies running the bots to care to do that.
If you do care bypassing Anubis is trivial.
Anubis is active when a user agent looks like a web browser (e.g. contains the "Mozilla" substring every major browser uses). The reverse proxy serves an interstitial page that does a proof-of-work check, validated server side, setting a cookie if it passes.
This means a legitimate user won't constantly get the proof of work check, because they already passed it. But AI bots rotating through tons of residential IPs to scrape your forum or git forge or whatever will be slowed down.
Overall, I like the idea. It's unobtrusive, privacy preserving, and seems to be working out well for a lot of sites.