Anthropic has announced it will ban users from engaging in "sustained and needless" abusive behaviour towards its AI systems, a move that has sparked considerable debate about how people should interact with artificial intelligence tools.
The maker of Claude said on Thursday it was updating its usage policy to allow its tools to end interactions where people are "cruel" — something it said it had done in "rare" cases previously. According to Anthropic's announcement, the updated policy takes effect on November 12, 2026.
The restriction now appears alongside other prohibited conduct including bullying others, promoting self-harm and creating non-consensual intimate imagery. However, Anthropic clarified that the policy would only apply in "extreme cases" of repeated abuse, not to "common versions of user frustration, pushback, dark creative themes, or model testing and research."
The company has not specified whether enforcement would extend beyond ending conversations, such as permanently banning users. According to reporting on the policy update, Anthropic did not clarify what additional measures, if any, might accompany conversation termination.
What counts as abusive behaviour?
The precise definition of what Anthropic considers "abusive or cruel behaviour" towards its models remains ambiguous. Screenshots of the new policy have circulated widely on social media, generating mixed reactions. Some commenters have praised the approach as advancing good manners or promoting what they describe as "model welfare."
Others have criticised the initiative sharply. One commenter called it "deeply wrong," while another suggested the focus on AI treatment might diminish concern for abuse directed at humans and animals. Dr Barry Scannell, technology partner at Irish law firm William Fry, expressed scepticism on LinkedIn, stating:
"You can't be cruel to numbers and maths. This level of anthropomorphisation of AI is harmful. It leads people to believe that it's something it's not."
Microsoft AI director Mustafa Suleyman, who criticised Anthropic in September for treating AI as if it were human, responded to the update with a melting face emoji on social media.
How does this fit into broader policy changes?
The update emerged from Anthropic's annual review of its usage policy and addressed multiple emerging risks beyond user conduct. The company also updated restrictions to prevent Claude from being used for "deceptive campaigns" and clarified that it cannot be deployed to "deceive voters or disrupt elections" as US midterm elections approach.
Anthropic also reiterated that its usage policy previously prevented its tools from being used to develop weapons. According to the company's policy update, the revision also clarifies restrictions on high-risk health and financial uses, autonomous physical actions, weapons-related software and components, and surveillance.
Despite these multiple changes, the restriction on abusive user behaviour has attracted far more attention and scrutiny than any other element of the update.
Should people be polite to AI tools?
The policy has reignited a broader debate about how users should communicate with generative AI systems. Some experts argue that writing or saying "please" or "thank you" to tools like ChatGPT is unnecessary and wastes tokens — the units that large language models use to process and generate responses.
Others contend that politeness can lead to higher quality answers or more helpful responses. When a social media user asked about electricity costs incurred by ChatGPT processing polite messages, OpenAI boss Sam Altman responded that it was "tens of millions of dollars well spent — you never know."
The Collins Dictionary highlighted the re-emergence of the term "clanker" for disparagingly referring to AI tools. Originally used to describe robots in Star Wars games and movies since the mid-2000s, the term made the dictionary's shortlist of words reflecting the mood of 2025 after July 2025 saw many people share frustrations with AI-powered machines on TikTok.
Could rudeness to AI affect human behaviour?
Some experts have suggested that being rude to AI chatbots could negatively influence how people communicate with and treat others. The concern is that hostile interactions with AI systems might normalise discourteous behaviour that could seep into human relationships and social interactions.
Anthropic's research into what the company describes as "model welfare" has been ongoing. According to reporting on the development, Anthropic had previously characterised its work on Claude's ability to end abusive conversations as part of this broader research initiative. The company confirmed that Claude's ability to end persistently abusive conversations was already operational on Claude.ai and Claude Code before the new policy was formally announced.
According to Anthropic's statement, ending conversations will remain the primary enforcement mechanism for abusive interactions going forward.
What happens next?
The revised usage policy is scheduled to take effect on November 12, 2026. Users and developers will need to familiarise themselves with the updated restrictions and the company's approach to enforcing them.
