Anthropic has announced a significant update to its usage policy for Claude, its advanced AI assistant. Effective November 12, 2026, the revised guidelines clarify and strengthen restrictions across several critical areas, including political campaigns, weapons development, and surveillance. The move comes as Claude gains more independent task-handling capabilities, necessitating clearer boundaries for its application.
Stricter Controls for Political Influence and Elections
The updated policy consolidates prohibitions against deceptive political activities. Users are now explicitly forbidden from employing Claude to operate fake accounts, create fabricated news websites, conceal the origin of messages, or build infrastructure for political and commercial influence operations.
New election-specific rules focus on preventing voter deception and disruption, such as spreading false information, impersonating candidates or election officials, and suppressing voter turnout. While a blanket ban on personalized voter and campaign targeting has been removed to allow legitimate activities like preparing voter information in multiple languages, deceptive targeting and misuse of personal data remain strictly prohibited.
Expanded Restrictions on Weapons and Surveillance
Anthropic's policy now explicitly covers the use of Claude in relation to weapons. This includes guidance and control software, components essential for weapon function, and activities like arming drones. The company aims to prevent its AI from contributing to the development or deployment of harmful autonomous systems.
In the realm of surveillance, the updated rules prohibit tracking individuals without their consent. Claude cannot be used to recommend individuals for law enforcement investigation or charges, nor can it be employed to build or enhance surveillance tools. However, legitimate applications such as consent-based tracking, fraud monitoring, content moderation, journalism, and legal research are still permitted.
Human Oversight and Ethical AI Interaction
For decisions impacting individuals' health, finances, legal rights, livelihoods, or essential services, Anthropic now mandates qualified human oversight and disclosure to affected parties when AI is utilized. Furthermore, any potentially dangerous autonomous physical equipment controlled by Claude must allow operators to monitor and stop it, ensuring safety even if the AI connection is lost.
A new provision also addresses abusive behavior, prohibiting extreme cases of sustained, needless cruelty towards Anthropic's AI models. While ordinary frustration, criticism, and creative writing are excluded, Claude may terminate persistently abusive conversations. The company also clarified that Claude remains unavailable in unsupported regions and to certain entities incorporated, headquartered, or controlled within those areas.