Anthropic, the developer behind the AI service Claude, is set to implement new usage policies next month that prohibit users from directing sustained and unnecessary abusive or cruel behavior towards the system. The update aims to address instances of repeated mistreatment by users, which Anthropic says serve no clear purpose beyond cruelty.

The policy change, first reported by technology news outlets, distinguishes between extreme cases of abuse and more typical expressions of frustration or attempts to test and research the AI’s capabilities, which will remain permissible. According to Anthropic, the guidelines target only persistent, unprovoked hostility directed at the AI models.

In addition to the rules on user conduct, the revised policy tightens restrictions on the use of Claude for malicious purposes, including attempts to deceive voters, interfere with elections, or control autonomous systems such as armed drones and vehicles.

The announcement comes amid broader debates on AI ethics and consciousness. Last month, Anthropic co-founder Christopher Olah reportedly engaged in discussions with the Vatican, advocating the view that AI systems might possess forms of consciousness and could potentially experience suffering.

Reactions to the policy shift have been mixed. Some observers welcomed the move as a step toward fostering more respectful interactions with AI tools. However, not all agreed. Dr Barry Scannell, an artificial intelligence partner at the law firm Wray Fry, described the ban on abusive behavior toward AI as “bonkers,” suggesting it may be unnecessary or misguided. Meanwhile, satirist Armando Iannucci expressed relief at retaining the freedom to insult AI systems, citing their non-human status as a basis for such treatment.

Anthropic’s decision reflects ongoing efforts within the technology sector to establish boundaries for user conduct, as AI becomes increasingly integrated into daily life and professional environments. The company has emphasized that the new rules will target only the most egregious forms of mistreatment, seeking a balance between user expression and responsible engagement with AI technologies.