Anthropic has updated its usage policy, the rulebook for what you can and can't do with Claude, and the new version takes effect on November 12. For most people the honest summary is short: nothing you do today will change. But a few lines are worth knowing about, including a new rule about how people treat Claude itself.
What actually changed
Anthropic published the update on October 8, 2026. It says most of the changes clarify rules that already existed. Three are worth a look.
- A new rule against cruelty toward Claude. The policy now prohibits "sustained and needless abusive or cruel behavior" toward its models. Anthropic says this applies only in extreme cases, where someone is repeatedly cruel for no apparent purpose.
- Voting help is less restricted. The old blanket ban on personalized vote and campaign targeting is gone. That means legitimate civic work, such as sharing voter information in other languages, is no longer lumped in with it. Deceptive targeting and misuse of personal data are still not allowed.
- Clearer rules for high-stakes uses. Where Claude is used to help make decisions about health, finance or legal rights, the policy still calls for a qualified human in the loop and notice to the person affected. The update spells out more clearly which kinds of recommendations are covered.
There is also a clarification about supported regions. Anthropic says the underlying restriction hasn't changed.
What it means for you at work and at home
If you use Claude to write emails, understand a letter, plan a trip or sort out a spreadsheet, you will not notice any difference. The cruelty rule is not about being short with Claude, typing in a hurry, or saying "no, that's wrong, try again."
Anthropic says it does not cover ordinary frustration, pushback, dark themes in creative writing, or testing and research. So a story with a villain in it is fine. Telling Claude it got something wrong is fine.
If you work somewhere that uses Claude to help make decisions about people, such as who gets a service or how a claim is handled, the high-stakes language matters more. The policy expects a real person to be involved and the affected person to be told. If that describes your job, it's worth reading that section of the policy yourself, or asking whoever set Claude up at your workplace.
The catch, and what's unclear
The main way Anthropic says it will handle the cruelty rule is by ending a conversation that has turned persistently abusive, on Claude.ai and in Claude Code. The policy page doesn't spell out other penalties, so I can't tell you what happens beyond that. And "extreme" and "needless" are judgment calls, so exactly where the line sits isn't something I can map out for you.
Some of the headlines about this change made it sound as if being rude could get you banned right away. Anthropic's own wording is much narrower than that.
What to do now
Nothing. There's no setting to change and nothing to sign. If you'd like to read it yourself, the policy is on Anthropic's website under its news section, titled "2026 Usage Policy update." It's written in fairly plain language, and the parts above are the ones most likely to touch everyday use.
If you'd like to see Claude used for everyday jobs on screen, I walk through them step by step on the YouTube channel @DoWithClaude.