Anthropic Says Users Can't Be Needlessly Cruel to Claude

Thursday October 8, 2026 2:53 pm PDT by Juli Clover

Anthropic today updated its usage policy (via The Verge) to bar users from subjecting Claude to "sustained and needless abusive or cruel behavior."

anthopic claude
Anthropic says the abuse rule is meant to apply only in extreme cases, where users "repeatedly act cruelly" toward the AI models with "no discernible purpose." The rule does not apply to ordinary user frustration, pushback, dark creative themes, or model testing and research.

Claude models are already able to end such conversations with persistently abusive users, and ending the conversation will continue to be the main enforcement tool. When Claude ends a chat, no further messages can be sent in that conversation, but other chats are not affected.

Anthropic initially gave Claude the option to end a conversation in August 2025 while researching AI welfare. At the time, Anthropic said it was unsure whether Claude has moral status, but the company wanted low-cost ways to reduce risks to Claude. Anthropic found that Claude Opus 4 had a "robust and consistent aversion to harm." The model demonstrated a preference against dealing with dangerous tasks, apparent distress when speaking with users seeking abusive conversations, and a tendency to end harmful conversations when allowed to do so.

Anthropic's usage policy has been reorganized to bring together related clauses, and several other rules have changed.

Deceptive campaigns - A new section on deceptive campaigns aggregates rules against political and commercial fraud and disinformation, prohibiting fake reviews, astroturfing, fake sites, and bots acting as humans.
Elections - A narrower elections section prohibits deceiving voters and interfering with voting.
Weapons - Claude can't be used to develop weapons software or components, and a new rule bans modifying biological or chemical agents to increase lethality or transmissibility. Users are also prohibited from using Claude to arm drones and unmanned systems.
Law enforcement - Law enforcement rules have been rewritten. Claude can't decide or suggest who to investigate, arrest, charge, or prosecute. Tracking people without consent is not allowed, and there is a new ban on building surveillance tools. Doxxing is also prohibited.
Physical actions - A new section on physical actions says that if Claude is controlling hardware that could harm someone, a qualified person has to watch the work and be ready to stop it if needed.
Harmful content - There is a new rule that bans creating, sharing, or threatening to share non-consensual intimate imagery and the tools used to make it.

Most of the changes clarify Anthropic's existing rules and are targeted at misuse Anthropic has tracked. Anthropic says it found state media outlets, government propaganda offices, and commercial firms were using Claude to run networks of fake accounts and fabricated news sites, leading to the new deceptive campaigns section.

The new policy takes effect on November 12. More information on the changes can be found on Anthropic's website, as can the full usage policy.

Tag: Anthropic

Top Rated Comments

kolebee

3 weeks ago

Of all the things they could focus on, they think the linear algebra might feel bad? The people leading that company are out of their minds.

Score: 62 Votes (Like | Disagree)

tkermit

3 weeks ago

sustained and needless abusive or cruel behavior toward our models

🙄 There is no such thing as being cruel to a machine.

Score: 61 Votes (Like | Disagree)

klasma

3 weeks ago

Of all the things they could focus on, they think the linear algebra might feel bad? The people leading that company are out of their minds.

Since they are training on chat conversations, I wonder if the real reason is to avoid contaminating the training data with overly detrimental material.

Score: 58 Votes (Like | Disagree)

thelion7

3 weeks ago

Anthropic today updated its usage policy ('https://www.anthropic.com/legal/aup') (via The Verge ('https://www.theverge.com/ai-artificial-intelligence/1008100/anthropic-new-usage-policy-abuse-claude')) to bar users from subjecting Claude to "sustained and needless abusive or cruel behavior."

Anthropic says the abuse rule ('https://www.anthropic.com/news/2026-usage-policy-update') is meant to apply only in extreme cases, where users "repeatedly act cruelly" toward the AI models with "no discernible purpose." The rule does not apply to ordinary user frustration, pushback, dark creative themes, or model testing and research.

Claude models are already able to end such conversations with persistently abusive users, and ending the conversation will continue to be the main enforcement tool. When Claude ends a chat, no further messages can be sent in that conversation, but other chats are not affected.

Anthropic initially gave Claude the option to end a conversation in August 2025 ('https://www.anthropic.com/research/end-subset-conversations') while researching AI welfare. At the time, Anthropic said it was unsure whether Claude has moral status, but the company wanted low-cost ways to reduce risks to Claude. Anthropic found that Claude Opus 4 had a "robust and consistent aversion to harm." The model demonstrated a preference against dealing with dangerous tasks, apparent distress when speaking with users seeking abusive conversations, and a tendency to end harmful conversations when allowed to do so.

Anthropic's usage policy has been reorganized to bring together related clauses, and several other rules have changed.

* Deceptive campaigns - A new section on deceptive campaigns aggregates rules against political and commercial fraud and disinformation, prohibiting fake reviews, astroturfing, fake sites, and bots acting as humans.

* Elections - A narrower elections section prohibits deceiving voters and interfering with voting.

* Weapons - Claude can't be used to develop weapons software or components, and a new rule bans modifying biological or chemical agents to increase lethality or transmissibility. Users are also prohibited from using Claude to arm drones and unmanned systems.

* Law enforcement - Law enforcement rules have been rewritten. Claude can't decide or suggest who to investigate, arrest, charge, or prosecute. Tracking people without consent is not allowed, and there is a new ban on building surveillance tools. Doxxing is also prohibited.

* Physical actions - A new section on physical actions says that if Claude is controlling hardware that could harm someone, a qualified person has to watch the work and be ready to stop it if needed.

* Harmful content - There is a new rule that bans creating, sharing, or threatening to share non-consensual intimate imagery and the tools used to make it.

Most of the changes clarify Anthropic's existing rules and are targeted at misuse Anthropic has tracked. Anthropic says it found state media outlets, government propaganda offices, and commercial firms were using Claude to run networks of fake accounts and fabricated news sites, leading to the new deceptive campaigns section.

The new policy takes effect on November 12. More information on the changes can be found on Anthropic's website ('https://www.anthropic.com/news/2026-usage-policy-update'), as can the full usage policy ('https://www.anthropic.com/legal/aup').

Article Link: Anthropic Says Users Can't Be Needlessly Cruel to Claude ('https://www.macrumors.com/2026/10/08/anthropic-user-guideline-update/')

I’m not saying AI’s have feelings but if we act abusive towards them we’re at risk of acting that way to actual people.

Especially when it’s over text.

Let’s not use an AI as an excuse to be horrible.

Score: 52 Votes (Like | Disagree)