The Pulse
Anthropic Rewrites Claude Rules on Influence Campaigns and Weapons
Anthropic published a revised Claude Usage Policy on October 8, 2026, spelling out bans on deceptive campaigns and weapons-related software. The policy takes effect November 12 and also narrows election restrictions while clarifying rules f
AI.info Team · Written with AI ·
“Most of the updates in the latest version are intended to clarify existing rules.”
Anthropic, 2026 Usage Policy update
Anthropic published a revised Usage Policy for Claude on October 8, spelling out restrictions on deceptive campaigns, weapons-related software and surveillance. The company says many changes make existing rules more explicit as Claude handles longer, more independent work. The new policy takes effect November 12.
Fake accounts and campaign infrastructure get a dedicated ban
The update gathers restrictions on deceptive activity into a new section covering political and commercial campaigns. It bars using Claude to conceal who is behind a message, amplify content through fake accounts or posts, or build tools and infrastructure for influence operations. Anthropic says the previous policy already prohibited the conduct, but its rules were spread across sections on elections, fraud, privacy and disinformation.
The company points to patterns in its September threat intelligence report, which describes operations involving fake accounts and fabricated news sites. In one case, Anthropic says a platform built with Claude managed more than 1,000 fake accounts and a fabricated news site called Malaysia Pulse. The report also describes a request for one million artificial views; Anthropic cautions that figures recorded by the actors’ own dashboard could not be independently verified.
Anthropic removes a blanket rule on voter targeting
The election section now focuses on conduct intended to deceive voters or disrupt elections, including false information about candidates or voting, impersonation of election officials, and efforts to suppress turnout. Anthropic renamed it “Do Not Undermine Democratic Processes.”
At the same time, the company removed its blanket prohibition on personalized vote and campaign targeting. Anthropic says that rule had also covered legitimate civic work, such as nonprofits preparing voter information in other languages or election officials sending ballot-cure notices. Targeting that relies on deception or misuses personal data remains prohibited under other sections.
Weapons rules now name guidance software and drone arming
Anthropic says it has seen attempts to use Claude to develop weapons guidance and control software. The revised policy specifies that the weapons ban includes the software and components that make weapons function, as well as arming drones and other autonomous vehicles.
The September report supplies a concrete example of the concern behind that wording. Anthropic says actors used Claude Code to work on guidance, navigation and control software for a guided rocket, including integrating an autopilot on a phone-class flight computer and running a flight simulation. The company says it does not have evidence that the actors fielded an operational device, though it reports that they test-fired a guided rocket and that the test appeared to fail.
Surveillance, physical systems and model abuse also change
The rewritten surveillance rules prohibit tracking people without consent, whether in real time or through analysis of previously collected data. Claude also cannot be used to recommend whom law enforcement should investigate, arrest or charge, or to build or improve surveillance tools. Anthropic says permitted uses include agreed-to tracking, content moderation, journalism and legal research.
Other changes clarify safeguards for high-risk uses and systems connected to equipment that can act physically. A qualified operator must be able to observe and stop such equipment, and it must hold a safe state if Claude disconnects. The policy also prohibits sustained, needless cruelty toward the models in extreme cases, while explicitly excluding ordinary frustration, model testing, research and dark creative themes.
The revision both tightens and narrows the written rules: it names more specific routes to harm while lifting a broad restriction that had swept in some civic work. Anthropic says the weapons and surveillance clarifications do not change what it enforces in practice. The revised policy is scheduled to take effect November 12, 2026.