Anthropic rewrote its Claude usage policy on weapons and surveillance: Reveneau

Why it mattersA team building any product on Claude has until 12 November to read the new policy and decide whether their current prompts, tools, and allowed outputs still fit, because the clauses on weapons software, surveillance, and autonomous physical actions each name specific product shapes rather than general categories.

A product that ships on Claude has a deadline of 12 November to be checked against the clauses Anthropic has just rewritten.

Anthropic published its 2026 Usage Policy update and said the new policy takes effect on 12 November 2026. The company says most of the changes clarify what the earlier policy already covered, but it also adds a prohibition on sustained abusive behaviour toward Claude and tightens the language on weapons, surveillance, and autonomous physical actions.

Weapons now explicitly covers software

The weapons section used to prohibit using Claude to develop weapons. Anthropic says the new version makes clear that the prohibition includes the software and components that make weapons work, as well as actions like arming drones and other autonomous vehicles. The company says this is how it already enforced the old policy, and the rewrite only makes the text match the enforcement.

Surveillance and criminal justice get narrower limits

Anthropic rewrote the surveillance and criminal justice section after its September threat intelligence report described AI being used to build systems that identify and track political dissidents. The new text prohibits tracking people without their consent, whether in real time or by analysing previously collected data, and prohibits Claude from deciding or recommending who to investigate, arrest, or charge. It also prohibits using Claude to build or improve surveillance tools. Anthropic names the uses that remain permitted: tracking people have agreed to, fraud monitoring, content moderation, journalism, and legal research.

Autonomous physical actions need a human who can stop them

The high-risk use case section now covers Claude when it is connected to hardware that takes autonomous physical actions and could cause injury. Anthropic says a qualified operator must be able to observe the equipment and stop it if needed, and that the equipment must hold a safe state if Claude is disconnected. Earlier high-risk rules on health, finance, legal rights, livelihood, and essential services are unchanged, and still require a qualified human in the loop and a disclosure to the affected person.

Elections and deceptive campaigns

Anthropic consolidated existing rules on fake accounts, impersonation, and influence operations into a new section called Do Not Engage in Deceptive Campaigns or Artificial Activity. The elections section was renamed Do Not Undermine Democratic Processes and now focuses on deceiving voters or disrupting elections. The company removed its blanket ban on personalised vote and campaign targeting, saying legitimate civic work such as translating voter information or sending ballot cure notices is now permitted, while targeting that relies on deception or on misusing voters' personal data remains prohibited under the deceptive campaigns, surveillance, and privacy sections.

A separate clause now prohibits users from being sustained and needlessly cruel to Claude in ways with no discernible purpose. Anthropic says it is meant to apply only in extreme cases.

Source

2026 Usage Policy update, Anthropic
Usage Policy, Anthropic

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.