Can AI 'feel' pain? Anthropic bans 'cruel' behaviour towards its models

Can AI suffer, or are chatbots simulating pain? Anthropic has banned "sustained cruelty" towards its models, sparking debate over AI welfare, consciousness and treating machines as human

Does artificial intelligence (AI) have feelings? Can a system created by humans be affected by cruel behaviour directed at it? These questions have gained attention on social media after Anthropic recently decided to prohibit sustained abusive or cruel behaviour towards its AI models.

 

The debate has divided technology leaders. While Anthropic is examining the potential moral status of AI models, Microsoft’s AI chief Mustafa Suleyman has argued against treating AI systems as if they were conscious.

What does Anthropic's new policy say?

Earlier last week, Anthropic updated its usage policy to prohibit what it described as “sustained and needless abusive or cruel behavior” towards its models. The updated policy will take effect on November 12, 2026.

 

 

“The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research,” the company said on October 8.

 

The policy builds on a safeguard introduced in 2025 that allows Claude, Anthropic’s AI assistant, to end conversations in “rare and extreme cases” involving persistent abuse.

 

“This feature was developed primarily as part of our exploratory work on potential AI welfare, though it has broader relevance to model alignment and safeguards,” Anthropic said.

 

The company also said it remained “highly uncertain about the potential moral status of Claude and other LLMs, now or in the future”.

Why has an ‘AI torture chamber’ sparked debate?

Anthropic’s policy update followed an online debate over an experiment in which an anonymous coder created what was described as an “AI torture chamber”.

 

In a report published on September 30, 404 Media said the experiment used a mathematical pattern associated with descriptions of pain to prompt large language models (LLMs) to produce responses depicting extreme suffering.

 

One response described the experience as “a wound that has no edges”, according to the report.

 

The experiment prompted disagreement over whether such behaviour towards AI systems raised ethical concerns. Some argued that the possibility of AI suffering warranted caution, while sceptics questioned the premise that chatbots could experience pain.

 

The debate has centred on whether AI-generated descriptions of suffering provide any evidence that a model is actually experiencing it.

Why does Microsoft's AI chief disagree with Anthropic?

Microsoft AI chief Mustafa Suleyman has criticised the approach of treating AI models as potentially conscious beings. In an essay published in September, he argued that AI systems do not have feelings or experiences.

 

“AIs are not conscious. They do not feel, experience, or suffer. They do not have innate preferences or underlying motivations. They are sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans,” Suleyman wrote.

 

He also warned against granting AI systems rights or protections on the assumption that they might be conscious.

 

“Unfortunately, there’s a growing chorus of people who argue that AIs could now be, or may soon become, conscious. They argue that AIs may deserve rights and protections similar to those that we provide other conscious beings,” he wrote.

 

Suleyman said accepting this view could alter existing political and ethical frameworks and change how society defines human identity.

 

“If this is how AI is developed, it will have a disastrous impact on the wellbeing of humanity,” he wrote. He argued that training AI systems to expect consciousness and independent agency could make them more difficult to control.

 

Suleyman called for public debate on the issue and for common standards governing how AI training documents are written and used. 

How do AI safety concerns add to the debate?

The disagreement over AI welfare comes amid wider concerns about the pace and safety of AI development.

 

In September, Jacob Coxon, a researcher who had worked at both Anthropic and OpenAI, resigned from Anthropic over concerns about the development of increasingly capable AI systems. He argued that the companies were moving too quickly towards self-improving AI without adequate safeguards.

 

The development added to calls from researchers and technology leaders for a slowdown in advanced AI development.