Anthropic's new Claude rules: what changes on 12 November, including a ban on abusing the AI
Anthropic has updated Claude's usage policy, from weapons and surveillance to a new rule against sustained, needless cruelty toward the model. What's changing.
Stated by the maker, publisher, or platform holder.
Anthropic has rewritten the rules for using Claude. The update was published on 8 October and takes effect on 12 November. Most of it tightens or clarifies what people can't use Claude for. One addition is unusual: a rule against being deliberately cruel to the AI itself.
Everything below is from Anthropic's announcement and its full usage policy.
The rule getting the attention
Coverage of the update, including MacRumors, notes that ordinary frustration, pushing back, dark creative themes and testing the model aren't the target. In practice, if Claude ends a chat, you can't continue that conversation, but your other chats aren't affected.
The other changes
What it means for you
If you use Claude normally, almost nothing changes. The new cruelty rule is aimed at repeated, pointless abuse, and its main effect is the one that already exists: Claude can end the conversation.
If you build on Claude, read the full policy before 12 November. The weapons, surveillance and physical-action sections are more specific than before, and the high-risk section now spells out which uses need a human in the loop.
One change loosens the rules rather than tightening them. Dropping the blanket ban on personalised vote and campaign targeting means some political targeting that was banned outright is now judged under the narrower "democratic processes" rules instead.
Why the cruelty rule exists
Anthropic has been publicly exploring whether its models could have something like wellbeing, while saying it's uncertain whether they do. Letting Claude end abusive conversations in 2025 was the first practical step, and this rule writes the same idea into the policy.
It comes in a busy month for Anthropic, which is also one of the companies in the FTC's investigation into AI agent risks and is reportedly preparing to go public.