1300 166 138 Tell us what you need
News / Anthropic

Getting angry at Claude is fine, but cruelty will be banned

By Monika Tantau • Published 1 month ago
Frustrated office worker gesturing angrily at his laptop in a meeting room while colleagues look on, illustrating everyday anger at an AI chatbot like Claude

Key takeaways

• Anthropic's updated usage policy bans sustained and needless cruelty towards Claude from 12 November
• Frustration, pushback, dark creative themes and model testing are all still allowed
• Claude ending the conversation remains the main way the rule is enforced
• The same update tightens rules on fake accounts, weapons, surveillance and hardware

Most Read

Swearing at Claude after a bad answer will not get you in trouble. Treating it cruelly, again and again, for no reason, soon will. Anthropic has added a ban on sustained and needless abusive or cruel behaviour towards its models to its usage policy, which takes effect on 12 November, according to Anthropic.

It is an unusual step: a rule about how people treat the chatbot itself, written into the terms users agree to. The cruelty clause is also only one part of a broader update, and the rest of it matters more to most businesses.

What counts as cruelty to Claude

Anthropic says the rule is meant to apply only in extreme cases, where a user is cruel to the model with no discernible purpose. It spells out what is not covered: common user frustration, pushback, dark creative themes, and model testing and research.

In practice, telling Claude it got something badly wrong, arguing with it or asking it to write a violent thriller is still fine. What the rule targets is a pattern of abuse that serves no task at all.

How Anthropic will enforce it

Anthropic does not describe account bans for this rule. It says Claude’s ability to end conversations with persistently abusive users on Claude.ai and in Claude Code will remain the primary enforcement mechanism. Once Claude ends a chat, no more messages can be sent in it, but the user can start a new one.

That ability dates back to August 2025, when Anthropic let Claude Opus 4 and 4.1 end a rare subset of conversations as part of its research into model welfare. At the time, the company said it was unsure whether Claude has moral status, but wanted low cost ways to reduce the risk in case it does.

Woman at a laptop reading through printed pages in a library, reviewing the fine print of Anthropic's updated usage policy for Claude

The rest of the policy update

The cruelty clause made the headlines, but most of the changes deal with misuse by people. A new section bans deceptive campaigns and artificial activity, such as fake accounts, fabricated news sites and influence operations, whether political or commercial. The elections section now focuses on deceiving voters and disrupting voting.

Weapons rules now cover the software and components that make weapons work, including arming drones. Claude cannot be used to decide or recommend who to investigate, arrest or charge, and tracking people without consent is banned. Hardware that Claude controls and that could hurt someone must have a qualified operator able to watch it and stop it.

Why Australian businesses should care

Every business that gives staff Claude accounts or builds on Anthropic’s models signs up to this policy, so the changes on 12 November apply to Australian teams too. The cruelty rule will rarely matter at work. The rules on fake reviews, surveillance and automated physical actions are more likely to, especially for marketing teams and for firms connecting AI to equipment.

This is a good moment to check that your internal AI use policy matches what your vendors now allow. Fake reviews are already a problem under Australian Consumer Law, and any AI that helps decide who gets investigated sits close to Privacy Act obligations.

Mapping those overlaps between vendor rules, local law and your own policy is a standard part of our AI consulting work.

A debate that is not going away

Not everyone agrees AI can be mistreated at all. Last month we covered how Microsoft and Anthropic disagree on whether AI has anything like feelings. Anthropic’s new rule does not settle that question, but it shows the company is willing to act as if the answer might be yes.

For most users nothing changes. Be as blunt with Claude as you like when it gets things wrong. Just know that from 12 November, endless cruelty for its own sake is against the rules.

Scroll to Top