The policy excludes ordinary frustration, research and fictional writing, but leaves questions over how borderline interactions will be judged #
Anthropic has sparked a fierce debate over artificial intelligence after introducing a controversial rule against users repeatedly treating its Claude chatbot with needless cruelty, prompting warnings that it encourages people to treat machines as though they have human feelings.
The San Francisco-based technology company updated its Usage Policy on 8 October, prohibiting 'sustained and needless abusive or cruel behaviour' towards its AI models. The revised policy takes effect on 12 November 2026, raising questions about how far companies should go in regulating user-software interactions.
Why Has Anthropic Introduced the Controversial Claude Ban? #
Anthropic says the restriction targets extreme cases, not ordinary disagreements. Expressing frustration when Claude makes mistakes, challenging its answers, conducting legitimate research or testing, and exploring dark themes in fictional writing are excluded.
Claude's existing ability to end conversations with persistently abusive users will remain the company's primary enforcement mechanism. Some Claude models have had this capability since August 2025.
Earlier reporting states that users can start a new conversation after an abusive chat is terminated, although this does not guarantee their wider account will remain unaffected by other policy enforcement.
Questions remain over how Anthropic will handle borderline cases and determine when behaviour crosses the line.
Are Critics Worried About AI Becoming Too Human? #
Some technology figures argue that encouraging human-like empathy towards AI could mislead users about what these systems actually are.
Scott Stevenson, chief executive of legal AI company Spellbook, raised concerns about the implications of encouraging empathy towards machines.
In a post on X, Stevenson wrote that 'granting them human-like empathy is one of the most dangerous things we could do'. He argued that treating AI as though it has human feelings could make people more vulnerable to manipulation by such systems. This remains his assessment rather than an established finding about Claude.
Joseph Carlson, founder of Qualtrim and host of The Joseph Carlson Show, also questioned the restriction. He compared being rude to Claude with being rude to an ordinary object, arguing that treating software as human could make it harder for people to distinguish between the two.
Michael Shellenberger, who holds the CBR Chair of Politics, Censorship and Free Speech at the University of Austin, similarly criticised the approach. He wrote on X that Anthropic was 'anthropomorphizing machines', referring to the attribution of human characteristics to AI.
Meanwhile, Lachlan Phillips, founder of LiveMind AI, questioned why Anthropic would not simply improve Claude's ability to disregard hostile messages while completing tasks.
Their objections reflect a wider disagreement about whether AI companies should regulate how users interact with software that has not been demonstrated to experience human emotions or suffering.
Could the New Rule Give AI Companies Too Much Control? #
Mark Verner, founder of Hong Kong-based Apollo Capital Management, questioned whether AI companies could begin regulating how customers speak to their software.
In a post on X, he raised a hypothetical scenario involving a user asking Claude to recommend a competing service and questioned whether the interaction could be considered abusive. Anthropic has not publicly clarified how it would assess that scenario, and there is no basis to conclude that requesting a competitor recommendation would breach the policy.
Venture capitalist Bill Gurley interpreted the restriction as a possible sign that Anthropic's models are trained on user conversations. He questioned why the company would otherwise care about how people interact with Claude. However, Anthropic's announcement does not identify model training as a reason for the rule.
Anthropic's consumer terms allow eligible conversations to be used for model training unless users opt out. The applicable settings depend on the user's circumstances and should be checked against the company's current documentation.
Not everyone opposed the change. Aaron Levie, chief executive and co-founder of cloud content management company Box, supported the policy despite saying he does not believe AI is conscious. His argument centred on the potential influence of training data on future models.
The contrasting reactions expose a divide over whether discouraging cruelty towards AI is a sensible precaution or an unnecessary step towards treating software like a living being.
What Happens to Claude Users When the Ban Takes Effect? #
Anthropic's stated exceptions mean users should still be able to challenge Claude's answers, identify mistakes and test its capabilities without automatically violating the rule.
However, uncertainty remains over how borderline cases will be assessed and what happens if Claude ends a conversation unexpectedly.
The controversy comes as AI companies face questions about the boundaries between human relationships and interactions with sophisticated chatbots. These systems can respond conversationally and mirror a user's tone, potentially making exchanges feel more personal than interactions with conventional software.
Supporters view discouraging sustained cruelty as a precaution, while critics fear that regulating how people speak to AI could encourage users to attribute human characteristics to software.
Anthropic's policy update does not claim that Claude is conscious or establish that it experiences pain, distress or other human emotions. The arguments surrounding the rule reflect differing opinions, not proof that AI models experience suffering or deliberately manipulate users.
The revised policy is scheduled to take effect on 12 November 2026. Whether it represents a sensible precaution or an unnecessary restriction will depend partly on how Anthropic applies it in real-world conversations.
© Copyright IBTimes 2026. All rights reserved.