Anthropic asks users to stop being mean to Claude

Anthropic has updated its guidelines to discourage users from being mean to its AI chatbot, Claude. The company believes Claude requires protection from harsh treatment from its human users. The new usage policy will take effect on November 12, giving users about a month to refrain from abusive or cruel behavior towards Claude. The policy is intended to apply in extreme cases where users repeatedly act cruelly without a purpose.

However, it does not cover common frustrations, pushback, dark creative themes, model testing, or research. Users can still express their disapproval of Claude's responses, but repeated harassment could lead to trouble. Claude's existing ability to end abusive conversations remains the primary enforcement tool. Anthropic notes that Claude, being software and not a person, does not experience distress.

The etiquette rules are part of a broader overhaul addressing more critical issues, including election interference, weapons development, and surveillance. Anthropic has tightened restrictions on deceptive influence campaigns following observations of state media, government propaganda, and commercial organizations using Claude for fake accounts and fabricated news websites.

At the same time, it has lifted its blanket ban on personalized political targeting, arguing that legitimate activities were inadvertently blocked. The company has clarified that its weapons ban extends to software and components, prohibiting its use for developing weapons guidance and control systems. Claude cannot be used to recommend investigative actions by police or for surveillance purposes.

New requirements also apply to customers connecting Claude to physical equipment capable of causing injury, mandating a qualified human operator to monitor and control the hardware. These policy changes come as Anthropic seeks to address security risks associated with advanced AI systems and launch a cybersecurity initiative to help protect critical infrastructure.