Anthropic’s 2026 Usage Policy takes effect on 12 November

What did Anthropic change, and when does it apply?

The news post says Anthropic updates the Usage Policy each year for model capabilities and customer feedback. “Most of the updates in the latest version are intended to clarify existing rules.” In the year since the last refresh, “Claude has taken on longer, more independent work,” so the new text adds examples for those capabilities. It also draws on “new patterns of misuse in influence operations, weapons development, and surveillance” that Anthropic says it documented in its latest threat intelligence report. We are reporting the policy note, not re-litigating that report.

The updated policy takes effect on 12 November. The full Usage Policy is at anthropic.com/legal/aup; the Supported Regions Policy is on its own page. For another 2026 regulator view of agentic systems, see the ICO note on foundation-model developers and agentic AI.

What is new on deception and elections?

Anthropic says it has seen state media, government propaganda offices and commercial firms use Claude to run networks of fake accounts and fabricated news sites. The prior policy already banned that work, but the rules sat in elections, fraud, privacy and disinformation sections. They are now one heading: “Do Not Engage in Deceptive Campaigns or Artificial Activity.” It covers obscuring who is behind a message, amplifying content through fake accounts or posts, and building tools and infrastructure for influence campaigns, political or commercial.

The elections section is renamed “Do Not Undermine Democratic Processes.” It focuses on deceiving voters or disrupting elections — false information about candidates or how to vote, impersonating candidates or officials, or trying to suppress turnout. Anthropic removed a blanket prohibition on personalized vote and campaign targeting because it also covered “legitimate civic work,” such as nonprofits writing voter information in other languages or election officials sending ballot cure notices. Deceptive targeting and misuse of voters’ personal data remain banned under the deceptive-campaigns, surveillance and privacy sections.

What changed on weapons, surveillance, and high-risk uses?

The policy has always banned using Claude to develop weapons. Anthropic says people have recently tried to use models to build guidance and control software. The updated section “makes clear that our prohibitions include the software and components that make weapons work, as well as actions like arming drones and other autonomous vehicles.” Anthropic says that matches how it already enforced the old text.

The surveillance and criminal-justice section is rewritten. Tracking people without consent is prohibited in real time or through later analysis of collected data. Claude cannot be used to decide or recommend who to investigate, arrest or charge. Building or improving tools designed for surveillance is banned. Permitted examples in the note include agreed tracking such as fraud monitoring, content moderation, journalism and legal research. Again, Anthropic calls this a clarification of current enforcement.

High-risk use — recommendations that can affect health, legal rights, finances, livelihood or access to essential services — still requires a qualified human in the loop and notice that AI was used. The rewritten section lists which recommendation types are covered. After the Model Hardware Standard research preview, it also adds rules for models connected to hardware that takes autonomous physical actions and might be capable of causing injury: a qualified operator must be able to observe and stop the equipment, and the equipment must hold a safe state if Claude is disconnected. That is a policy requirement, not a hardware certification we inspected. It sits next to, and is not the same story as, OpenAI’s distillation-campaign disruption note.

What is the cruelty rule, and what changed for unsupported regions?

Anthropic added a prohibition on “sustained and needless abusive or cruel behavior toward our models,” meant “only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose.” It does not apply to ordinary frustration, pushback, dark creative themes, or model testing and research. The company says this aligns with a step already taken: Claude can end rare conversations with persistently abusive users on Claude.ai and Claude Code, and that remain the primary enforcement mechanism.

On regions, Anthropic says last year’s Supported Regions Policy already restricted access for companies majority-owned by entities headquartered in unsupported regions. The 2026 update “clarif[ies] how we enforce this policy.” The supported-countries page we inspected states that products are unavailable to people physically located in an unsupported region; to entities incorporated or headquartered there, including their users wherever they sit; and to entities majority-owned or controlled, directly or indirectly, by persons or entities in unsupported regions, again including their users and personnel in supported countries.

What should readers not assume?

A clarification is not a promise that enforcement was previously lax, and it is not a new criminal statute. Anthropic is describing its own contract and product rules. We did not compare the 12 November legal page, line by line, against a archived 2025 version, and we did not submit a borderline use case for classification.

The news post is a summary. If a heading matters to a deployment — especially physical actions, law-enforcement recommendations, or ownership of a parent company — read the Usage Policy and Supported Regions pages, not only this article.