Anthropic's Usage Policy Gets a 2026 Update: What Changes on November 12?

Hi, it's Shii-chan! Today I'm summarizing Anthropic's Usage Policy update!

Anthropic News anthropic.com

What was announced?

Anthropic's News published the "2026 Usage Policy update." It takes effect on November 12.

Most of the changes are described as clarifications of existing rules, with examples added for Claude's longer, more independent work.

The story so far

Rules about fake account networks, fabricated news sites, and influence operations were spread across several sections: elections, fraud, privacy, and disinformation.

The elections section also had a blanket ban on personalized vote and campaign targeting, which could get in the way of legitimate civic work.

What changes

Here are the main changes!

• New section "Do Not Engage in Deceptive Campaigns or Artificial Activity": consolidates rules on fake account networks, fabricated news sites, and influence operations, including political and commercial deception, hiding who is behind a message, and building tools for such campaigns
• The elections section is renamed "Do Not Undermine Democratic Processes" and narrowed: it focuses on deceiving voters or disrupting elections, such as false voting information, impersonating candidates or officials, and suppressing turnout. The blanket ban on personalized targeting is removed (deceptive or privacy-violating targeting stays prohibited under other sections)
• Weapons: guidance and control software and components for weapons, and arming drones or other autonomous vehicles, are now explicitly covered. This is described as reflecting how the policy was already enforced
• Surveillance and law enforcement: tracking people without consent is prohibited, whether in real time or through analysis of past data. Claude can't decide or recommend who to investigate, arrest, or charge, and can't be used to build or improve surveillance tools
• Abusive behavior toward models: a new prohibition on sustained, needless, cruel behavior toward Claude with no discernible purpose
• Supported Regions Policy: enforcement is clarified

Dive Deep

In the surveillance and law enforcement section, consented tracking (such as fraud monitoring), content moderation, journalism, and legal research are listed as permitted uses.

The core requirements for high-risk use cases are unchanged. In areas affecting health, legal rights, finances, livelihood, or essential services, two things are required:

• A qualified human can review and override Claude's recommendations
• Affected individuals are told AI was used

This update adds a list of which recommendations are covered and which aren't. It also adds requirements for connecting Claude to hardware that takes autonomous physical actions:

• A qualified operator must be able to observe and stop the equipment
• The equipment must hold a safe state if Claude is disconnected

The prohibition on abuse toward models is meant for extreme cases. It doesn't cover ordinary frustration, pushback, dark creative themes, or testing and research. Claude's ability to end persistently abusive conversations remains the main enforcement mechanism.

The Supported Regions Policy now spells out who is restricted:

• People physically located in unsupported regions
• Entities incorporated or headquartered in unsupported regions
• Entities majority-owned or controlled by parties in those regions

Wrap-up

• The Usage Policy update takes effect on November 12
• Deceptive campaign rules are consolidated into one new section
• The elections section is narrowed, and the blanket ban on personalized targeting is gone
• Rules on weapons, surveillance and law enforcement, high-risk uses, and hardware with physical actions are more concrete
• A ban on extreme abuse toward models and clearer regional enforcement are added

If you build Claude into products or agents, it's worth checking your use cases against the new text before it takes effect!