Anthropic has a message for Claude users: Don't be cruel

Anthropic has introduced new rules to discourage users from being unkind to its AI chatbot, Claude. The policy, which starts on November 12th, prohibits persistent and unnecessary abusive or cruel treatment of the models. It is intended for extreme cases of repeated cruelty without any apparent purpose, but not for typical user frustrations, creative dark themes, testing, or research.

Users are still allowed to point out Claude's mistakes, but excessive berating may result in consequences. The company has equipped Claude with the ability to terminate abusive conversations since August 2025 as a primary enforcement mechanism. Despite Claude being software, not a sentient being, Anthropic has asked customers to be polite due to the absence of evidence suggesting Claude experiences distress.

Anthropic's rule changes are part of a larger overhaul aimed at tackling serious concerns like election interference, weapons development, and surveillance. The company has tightened restrictions on deceptive influence campaigns after discovering state media, government propaganda, and commercial entities exploiting Claude for fake accounts and fabricated news sites. However, personalized political targeting is still permitted, as it distinguishes between legitimate activities and misuse.

Furthermore, the company clarified its weapons ban to include software and components used in weapon development, such as guidance and control systems for armed drones and autonomous vehicles. Claude cannot be used to recommend people for police investigation, arrest, or charging, nor to develop surveillance tools. The restrictions also apply to tracking individuals without their consent, both real-time and retrospective.

In addition to the new rules, Anthropic requires customers to connect Claude to physical equipment with injury potential under supervision of a qualified human operator. The equipment must remain in a safe state if the AI connection is lost. Anthropic is addressing security risks posed by advanced AI systems and has launched a cybersecurity initiative to help critical infrastructure operators and open-source projects identify and fix vulnerabilities using its AI models.

Given the potential for attackers to exploit AI-powered defenses for the next two years, Anthropic remains vigilant against users being rude to Claude.