DDoS

July 20, 2025

People have liked talking to computers for quite a while now. Sometimes, the computers even respond - from ELIZA, to, well, the AI homunculi over at janitorai (all lower-case, we're back to the mid-2000s). What is janitorai, one might ask. Well, to keep it brief, it's a collection of AI chatbots that have been fine-tuned to behave like various fictional and non-fictional characters. Want to talk to Goku? Check. Jesus? Check. Ami, "your personal slave girl"? Check. You get the gist of it, role-playing without the existential dread of imagining what the person at the other end of the screen looks like.

In a clearly related note, the UK has passed a potentially very consequential bill called the "Online Safety Act". I barely need to say anything about it, it is meant to protect children online, give adults more control over the content they see and, well, essentially bans a bunch of content including but not limited to CSAM, "fraud", "extreme sexual violence", "terrorism" etc. What is most concerning is that the OSA expects that platforms preempt such content from appearing on people's feeds, mail boxes, chat screens, what have you. Obviously there's a lot to be said about the practicality of such orders, especially considering things like end to end encryption. Furthermore, the UK has the ability to go after any company on earth over this and fine it 18 million pounds, or 10% of their worldwide revenue (!!!), whichever is larger. Also, the regulator for UK communications services, ofcom, actually has a questionnaire that you can use to see if the regulations apply to you. We have a strict no-links policy outside of one or two that we can explicitly control and/or search for (you never know who owns that domain in 10-20 years), but you can find it pretty easily. I am happy to announce that the OSA does not apply to lainwired dot net. UK visitors will not be blackholed out of existence, yet.

Now, to tie these notes together, there is a sub-section in the bill that very specifically targets "harmful algorithms". If your AI and/or algorithm shows "bad content", well, stop it! Or get fined. What happens when your AI model is an effective black box, and a lot of its allure can be potentially described as, well, "obscene"? Well, according to janitorai, you leave the UK and tell its users in a very thinly veiled manner that they can use VPNs and their accounts are still up, but you're not putting up with it.

Hackers calmly discuss the pros and cons of having access to textual porn of your choice. Some play devil's advocate, some say "well if you can't protect children, who's fault is that?", some talk about the red tape in the UK and EU. There are people claiming that janitorai is just "too cheap to follow the rules". One would expect that at least on a subject so obviously related to the Four Horsemen of the Infocalypse, on a website that boasts tech-savvy users of a certain age bracket, a near-consensus that is reasonable could be reached. But no, the discussion is actually so mind-bogglingly stupid that I personally had to give up partway.