Anthropic updated a rather strange rule yesterday.
From now on, you can't insult Claude repeatedly over a long period without good reason.
And it's now officially part of the usage policy.
But don't get the wrong idea.
You won't get banned just for insulting Claude once when it writes bad code.
Anthropic says ordinary complaints, pushback, dark-themed creative writing, and even model testing won't be affected.
What's mainly prohibited is repeated, extreme, malicious interaction with no clear purpose.
More interestingly, Claude was already able to end certain conversations where it was being continuously insulted.
Now the company has simply made that an official rule.
It takes effect on November 12.
AI usage policies used to tell you what you couldn't do with AI.
Now they're starting to regulate how you treat AI, too.
Source: Anthropic official website
https://www.anthropic.com/news/2026-usage-policy-update
From now on, you can't insult Claude repeatedly over a long period without good reason.
And it's now officially part of the usage policy.
But don't get the wrong idea.
You won't get banned just for insulting Claude once when it writes bad code.
Anthropic says ordinary complaints, pushback, dark-themed creative writing, and even model testing won't be affected.
What's mainly prohibited is repeated, extreme, malicious interaction with no clear purpose.
More interestingly, Claude was already able to end certain conversations where it was being continuously insulted.
Now the company has simply made that an official rule.
It takes effect on November 12.
AI usage policies used to tell you what you couldn't do with AI.
Now they're starting to regulate how you treat AI, too.
Source: Anthropic official website
https://www.anthropic.com/news/2026-usage-policy-update