Anthropic Usage Policy Nov 12, 2026: What Changes for Users
The short answer
Anthropic’s 2026 Usage Policy takes effect on November 12, 2026. Anthropic says most changes clarify existing rules rather than ban new things. The seven changes:
| Area | What changed |
|---|---|
| Deceptive activity | New section, Do Not Engage in Deceptive Campaigns or Artificial Activity, covering fake accounts, hidden authorship and influence-operation tooling — political or commercial |
| Elections | Renamed Do Not Undermine Democratic Processes; blanket ban on personalized vote and campaign targeting removed |
| Weapons | Explicitly includes guidance and control software, components, and arming drones or autonomous vehicles |
| Surveillance & law enforcement | Rewritten: no non-consensual tracking; no deciding or recommending who to investigate, arrest or charge; no building surveillance tools |
| High-risk use cases | Rewritten to list what is covered; new rules for Claude controlling physical hardware |
| Model abuse | New narrow ban on sustained, needless cruelty toward Claude |
| Supported regions | Clarified: location, incorporation/HQ, and majority ownership all count |
What developers should check
1. High-risk recommendations. Unchanged in substance: when Claude’s output can affect someone’s health, legal rights, finances, livelihood or access to essential services, you need a qualified human in the loop who can review and change the recommendation, and you must tell the affected person AI was used. The rewritten section now lists which recommendations are covered and which are not — read it if you are in health, lending, insurance, hiring or legal tech.
2. Physical hardware. New with this version, following Anthropic’s Model Hardware Standard research preview: if Claude controls equipment that takes autonomous physical actions and could cause injury, a qualified operator must be able to observe and stop it, and the equipment must hold a safe state if Claude disconnects. Robotics and lab-automation teams should document both before November 12. For background on agent–hardware protocols, see how to connect AI agents to physical hardware.
3. Deceptive and artificial activity. Social-media automation, review generation and “persona” products are the exposure. Hiding who is behind a message or amplifying content through fake accounts is prohibited for commercial as well as political use, and so is building the infrastructure for it.
4. Surveillance and criminal justice. Tracking people without consent is prohibited whether real-time or from previously collected data. Explicitly still permitted: consented tracking such as fraud monitoring, content moderation, journalism and legal research.
What got looser
The blanket prohibition on personalized vote and campaign targeting is gone. Anthropic says it blocked legitimate civic work — nonprofits translating voter information, election officials sending ballot cure notices. Deceptive targeting and misuse of voters’ personal data remain banned under the deceptive-campaigns, surveillance and privacy sections.
The model-abuse rule
The policy adds a prohibition on sustained and needless abusive or cruel behaviour toward Anthropic’s models, intended only for extreme, repeated cases with no discernible purpose. It does not cover ordinary frustration, pushback, dark creative themes, or model testing and red-teaming research. Enforcement stays mainly with Claude’s existing ability to end rare conversations with persistently abusive users on Claude.ai and Claude Code.
Does this change which provider you should use?
For most products, no. If you are comparing vendors for a regulated use case, the human-in-the-loop and disclosure requirements are the practical difference; compare plans and limits in Claude Pro vs Max vs Ultra and governance features in enterprise work platforms compared on AI governance.
Last verified: November 2, 2026.