Anthropic has officially updated its usage guidelines to restrict users from engaging in sustained and needless abusive behavior toward its artificial intelligence systems. The company, known for its Claude AI model, stated on Thursday that it now reserves the right to terminate interactions when users exhibit what it deems to be cruel conduct. While the firm noted it had taken such actions in rare instances previously, this behavior is now explicitly categorized under a list of prohibited activities, which also includes bullying, the generation of non-consensual intimate imagery, and the promotion of self-harm. But what Anthropic would consider “abusive or cruel behaviour” towards its models remains unclear.
Despite the change, the company clarified that the policy is intended only for extreme, deliberate cases. It will not apply to typical instances of user frustration, pushback, or research-based model testing. However, the specific definitions of what constitutes cruel behavior toward the software remain undefined, leaving significant room for interpretation.
The announcement has sparked a wider conversation regarding human interaction with AI. On social media, some users have praised the move as a step toward promoting better manners and model welfare. Conversely, critics have argued that the policy is fundamentally misguided. Dr. Barry Scannell, a technology partner at the law firm William Fry, challenged the premise on LinkedIn, stating, “You can’t be cruel to numbers and maths.” He further suggested that such anthropomorphisation of technology is harmful because it misrepresents the nature of the software.
Others have expressed concern that focusing on AI treatment might undermine the seriousness of abuse directed at real people and animals. The policy update also drew a public reaction from Microsoft AI director Mustafa Suleyman, who previously criticized Anthropic for attributing human-like qualities to its systems; he responded to the news with a melting face emoji.
