Claude's New Usage Policy: What to Review Before November 12

If your team uses Claude for work, somewhere in your company there's a deployment nobody has looked at since it was set up: a marketing tool that pulls customer data for targeting, a support bot that flags "suspicious" accounts, a pilot that connects Claude to a piece of physical equipment. None of that is unusual. What's new is that Anthropic just updated its Usage Policy, and the changes take effect November 12, 2026. If you're running Claude in production, this is worth a deliberate look before that date, not after.

What actually changed

The update, effective across Claude's apps, API and Claude Code, reorganizes several sections and tightens a few in ways that matter for businesses:

• Deceptive activity. Rules against fake account networks, fabricated news sites and influence operations are now grouped under one section, "Do Not Engage in Deceptive Campaigns or Artificial Activity." It explicitly covers commercial activity, not just political campaigns, and includes building the tools or infrastructure for such campaigns, not only running them.
• Elections, narrowed. The section is renamed "Do Not Undermine Democratic Processes" and now focuses specifically on deceiving voters or disrupting elections. The old blanket ban on personalized voter and campaign targeting is gone, but targeting that relies on deception or misuse of personal data is still prohibited elsewhere in the policy.
• Surveillance. Non-consensual tracking, whether real-time or using previously collected data, is prohibited. Claude also may not be used to decide or recommend who to investigate, arrest or charge, or to build or improve surveillance tooling. Consented tracking, like fraud monitoring, plus content moderation, journalism and legal research, remain permitted.
• Weapons. The ban now explicitly names software and components that make weapons work, and arming drones or other autonomous vehicles.
• Hardware controlling physical actions. New requirements apply specifically when Claude is connected to hardware that takes autonomous physical actions in the real world.
• Supported Regions. Enforcement is clarified to cover people physically located in unsupported regions, and entities incorporated, headquartered, or majority-owned or controlled by parties there.

None of this is dramatic if your use of Claude is ordinary business work. But "ordinary" is exactly the assumption worth checking, because a few common setups sit closer to these lines than most teams realize.

Five things to check this week

1. Any tool that targets people using their personal data

If marketing or growth uses Claude to help personalize outreach, that's not banned outright, but targeting that relies on deception or on personal data used outside what customers agreed to is. Pull up every Claude-powered targeting or personalization workflow and check what data it uses and how customers consented to it.

2. Any tool that monitors or tracks individuals

Fraud monitoring with proper consent is fine. A tool that watches employee activity, customer behavior, or any individual's movements without clear consent is now explicitly flagged, whether it uses live data or data collected earlier. If you have anything resembling a monitoring or surveillance use case, get your legal or compliance lead to review it against the new Surveillance section directly, not a summary of it.

3. Any high-risk recommendation Claude makes without a named human reviewer

If Claude is involved in decisions in health, legal, finance, employment or essential services, the policy's human-in-the-loop requirement hasn't changed, but the revision clarifies which recommendations it covers. Confirm, for each such workflow, that a qualified person actually reviews and can override the output, and that anyone affected is told AI was used. "Someone could review it if they wanted to" isn't the same as a reviewer who does.

4. Any integration where Claude controls physical equipment

If you've connected Claude, directly or through an agent, to anything that moves, heats, cuts, dispenses or otherwise acts in the physical world, the new hardware provisions apply to you specifically. An operator needs to be able to observe the equipment and stop it, and the equipment needs a safe default state if Claude disconnects. Few Indian businesses will have this, but manufacturing, logistics and facilities teams piloting physical agents should check now.

5. Where your company and your users actually are

The Supported Regions clarification widens who it reaches: not just users physically located in an unsupported region, but companies incorporated, headquartered, or majority-owned or controlled there. If your company has any offshore ownership structure or serves users across borders, confirm eligibility against Anthropic's Supported Regions Policy rather than assuming your current status still holds.

Make this a 20-minute exercise, not a project

You don't need a legal review of every Claude integration in your company. You need one person, probably whoever already owns your AI tools, to list every active Claude deployment and run it against the five checks above. For most teams doing straightforward work, drafting documents, summarizing calls, writing code, answering routine customer questions, the answer will be "no concerns, nothing to change." The point of doing this now is to find the one deployment that isn't ordinary before November 12 finds it for you.

This is also a good moment to fold Claude-specific policy checks into whatever governance habit your company already has. If you don't have one yet, our earlier checklist on closing India's AI governance gap is a useful starting point: one named owner per AI system, and a simple cadence for checking what it actually does.

The bottom line

A vendor updating its usage policy isn't newsworthy by itself. A fixed compliance deadline on a tool your team already depends on is. Spend twenty minutes this week walking your Claude deployments through these five checks, fix what needs fixing before November 12, and move on. If you're building out AI use across a larger organization and want this kind of review built into your rollout process from the start rather than bolted on after a policy change, that's the kind of work we help with through AI consulting.