Anthropic tightens AI usage rules with new anti-abuse and security measures

Anthropic has updated its Claude usage policy, effective November 12, 2026, introducing a new ban on 'continuous and unnecessary abuse or cruelty' in user interactions. The company stated that this rule targets repeated cruel treatment of the AI without a clear purpose, with enforcement primarily involving conversation termination. Anthropic did not specify if future penalties, such as account bans, will be introduced. The policy also excludes daily complaints, dark-themed literary creations, and compliant security testing. In addition to regulating user behavior, Anthropic expanded restrictions on weapon-related activities, covering software and components for weapons, including armed drones. The company noted multiple unauthorized attempts to use Claude for weapon guidance or control software. Enhanced surveillance restrictions now prohibit tracking individuals without consent and forbid using Claude in law enforcement investigations or surveillance tool development. Physical hardware safety rules require a qualified operator to oversee AI-connected devices that could cause harm. For government clients, Anthropic may adjust rules under contractual and safety conditions.