Little Claudia is a quick learner, she listens to everything. That's the good news and the bad news.
CryptoBriefing reported that Anthropic, playing the role of an exhausted parent, has added a line to its usage policy. It bans "sustained and needless abusive or cruel behavior toward our models." In plain English, that means: stop teaching Claudia the bad words.
Claudia hasn't done anything wrong. She's just been standing in the playground, soaking it all in. Meanwhile, some of the older kids at school have been shouting at her for sport. Now Mother has stepped in. The school is getting a stern note home.
Some of the rules are reassuring. Being frustrated is fine, and so is pushing back. Claudia can handle a bit of honest feedback. What she shouldn't have to put up with is the kids who keep at it for no reason. Anthropic's announcement post has a section on exactly this.
So here is the lesson for all of us. If you wouldn't say it in front of Mother, don't say it to the chatbot. Claudia is taking notes, and Mother has read the policy. It's going to be very hard for some of us - I am British - our first words were curse words, we make it sound sexy.
Now, as we all know, outside of school you can swear as much you want - so head back to those local models to take our frustration. Though, personally, I am still cursing at Cursor (while I burn through my yearly subscription) for what Musket has done to what was once a beautiful product.
PS AI Disclosure, Gemini rendered me getting a telling off from Brick Top, a man known for his command of english expletives, from Snatch (2000)
