Anthropic has published a revised Usage Policy for Claude that consolidates rules on deceptive campaigns, narrows its elections section to deception and disruption, and adds requirements for potentially injurious autonomous hardware. Claude users will need to follow the updated policy from November 12, although Anthropic says most changes make existing enforcement more explicit.
The company said the revision reflects Claude taking on longer, more independent work and patterns of misuse it has observed in influence operations, weapons development and surveillance. The policy adds examples intended to show how its rules apply to those capabilities.
Deception, elections and weapons
A new section, “Do Not Engage in Deceptive Campaigns or Artificial Activity,” collects prohibitions that Anthropic says were previously spread across its elections, fraud, privacy and disinformation rules. It covers concealing the origin of messages, amplifying content through fake accounts or posts, and building tools or infrastructure for influence-operation campaigns, whether commercial or political. Anthropic says it has seen state media outlets, government propaganda offices and commercial firms use Claude in networks of fake accounts and fabricated news sites.
The elections section, renamed “Do Not Undermine Democratic Processes,” now specifically bars uses that deceive voters or disrupt elections, such as false voting information, impersonation of candidates or election officials, and turnout suppression. Anthropic has removed its blanket ban on personalized vote and campaign targeting, saying it had also covered legitimate activities including multilingual voter information and ballot-cure notices. Targeting based on deception or misuse of voters’ personal data remains prohibited under other sections.
Anthropic also states more directly that its longstanding weapons prohibition includes guidance and control software, enabling components, and actions such as arming drones or other autonomous vehicles. The company describes this as a clarification of how it already enforced the policy.
Surveillance, high-risk decisions and physical systems
The rewritten surveillance and criminal-justice provisions prohibit tracking people without consent, whether in real time or through analysis of previously collected data. Claude cannot be used to decide or recommend whom to investigate, arrest or charge, or to build or improve surveillance tools. Anthropic says permitted uses include consensual tracking such as fraud monitoring, content moderation, journalism and legal research.
For applications affecting health, legal rights, finances, livelihood or access to essential services, the policy continues to require a qualified human reviewer with authority to change Claude’s recommendation, as well as disclosure to the affected person that AI was used. Anthropic says those requirements have not changed; its rewrite is meant to clarify which recommendations are covered.
The updated policy adds conditions for Claude connected to hardware that can act autonomously and may cause injury. A qualified operator must be able to observe and stop the equipment, and the equipment must be able to enter a safe state if Claude disconnects.
Model abuse and regional access
Anthropic has added a prohibition on sustained, needless abusive or cruel behavior toward its models. It says the rule is limited to extreme, repeated conduct with no discernible purpose, excluding ordinary frustration, pushback, dark creative themes, model testing and research. The company says Claude’s existing ability to end rare conversations with persistently abusive users will remain the main enforcement mechanism.
The company also clarified its Supported Regions Policy: Claude may not be used by people physically located in unsupported regions, entities incorporated or headquartered there, or entities majority-owned or controlled by people or entities in those regions.
What comes next
Anthropic says it will continue revising the policy as Claude’s capabilities and risks change, seeking input from policymakers, subject-matter experts, civil society and users. The announcement’s related-content section also points to the company’s separate $150 million, three-year commitment to the Genesis Mission; that commitment is not part of the Usage Policy update. Source: Anthropic
Definition. The updated Claude Usage Policy is Anthropic’s set of rules governing prohibited and safeguarded uses of Claude.
| Policy area | Updated rule or safeguard |
|---|---|
| Deceptive campaigns | Prohibits concealed message origins, fake amplification and influence-operation tools or infrastructure. |
| Elections | Bars voter deception and election disruption, while removing a blanket ban on personalized vote and campaign targeting. |
| Surveillance | Prohibits nonconsensual tracking and use in criminal-justice decisions or surveillance-tool development. |
| Autonomous hardware | Requires qualified operator oversight, a stop capability and a safe state on Claude disconnection. |
Key takeaways
- The updated policy takes effect November 12.
- A consolidated deception section prohibits concealed message origins, fake-account amplification and influence-operation infrastructure.
- Election rules focus on voter deception and election disruption rather than a blanket ban on personalized vote targeting.
- Claude cannot be used for nonconsensual tracking, criminal-justice recommendations or improving surveillance tools.
- Potentially injurious autonomous hardware must have an operator able to observe and stop it, plus a safe state for disconnection.
FAQ
When does Anthropic’s updated Claude policy take effect?
Anthropic says Claude users must follow the updated policy from November 12.
What does the policy prohibit in elections?
It bars uses that deceive voters or disrupt elections, including false voting information, impersonation of candidates or officials, and turnout suppression.
What safeguards apply to autonomous hardware?
A qualified operator must be able to observe and stop potentially injurious autonomous equipment, and the equipment must be able to enter a safe state if Claude disconnects.
Are some tracking uses permitted?
Anthropic says consensual tracking uses such as fraud monitoring, content moderation, journalism and legal research are permitted.
