Anthropic Updates Usage Policy to Ban Persistent Abuse of Claude and Restrict Weapon Use

Anthropic has updated its usage policy for Claude to explicitly prohibit persistent and unnecessary abuse or cruelty toward the AI model. The revision also tightens restrictions on using Claude for weapon control software, armed autonomous hardware, non-consensual surveillance, and criminal justice decision-making. This policy update introduces a novel precedent in AI governance by explicitly addressing abusive user behavior toward AI models alongside physical safety risks. It highlights growing industry concerns regarding the integration of LLMs into autonomous warfare, surveillance tools, and safety-critical hardware. Enforcement for the anti-cruelty policy primarily involves terminating the conversation session, though it excludes legitimate research, model testing, dark creative themes, and regular complaints. Furthermore, any deployment of Claude with autonomous physical hardware now requires a qualified human operator present to halt operations if needed.