Anthropic's new usage policy takes effect, with stricter rules for Claude in hardware and high-stakes decisions
The update announced in October is now in force. It consolidates rules on deceptive campaigns, drops the blanket ban on campaign targeting, and adds a narrow ban on abusing the model.
Anthropic's revised Usage Policy takes effect today, November 12, 2026. The company published the changes on October 8 and gave customers five weeks of notice. Starting now, the new text is what governs use of Claude across Anthropic's products and API.
Anthropic says it revises the policy every year. In its account, most of this version clarifies existing rules and adds examples for the longer, more independent work Claude now does as an agent. Some of it responds to misuse Anthropic documented in its September 2026 threat intelligence report, covering influence operations, weapons development and surveillance. The clause that got the most attention, a ban on extreme abuse of the model itself, is one of the smaller changes in practice.
What changes today
- A new section on deceptive campaigns. Rules against fake accounts, fabricated news outlets and hidden sponsorship used to be split across the elections, fraud, privacy and disinformation sections. They now sit together under "Do Not Engage in Deceptive Campaigns or Artificial Activity", which covers both political and commercial operations. It also prohibits building tools or infrastructure for those campaigns.
- Elections, narrowed. The section is renamed "Do Not Undermine Democratic Processes" and focuses on deceiving voters, impersonating candidates or officials, and suppressing turnout. Anthropic removed the blanket ban on personalized vote and campaign targeting. It says that ban also blocked legitimate civic work, such as nonprofits producing voter information in other languages. Targeting that relies on deception or misused personal data is still prohibited under other sections.
- Weapons. The ban now explicitly covers the software and components that make weapons work, and arming drones and other autonomous vehicles. Anthropic says these changes reflect how it already enforced the previous policy.
- Surveillance and law enforcement. The section is rewritten. Tracking people without their consent is prohibited, whether in real time or by analyzing data collected earlier. Claude cannot be used to decide or recommend who to investigate, arrest or charge. Consented uses such as fraud monitoring, content moderation, journalism and legal research remain allowed.
- Abuse of the model. A new line prohibits "sustained and needless abusive or cruel behavior toward our models." Anthropic says it targets only extreme cases of repeated cruelty with no discernible purpose. It does not cover ordinary frustration, pushback, dark creative themes, or model testing and research.
The part developers should read
For teams building on the API, the high-risk requirements matter more than the cruelty clause. The core rules are unchanged. If Claude produces a recommendation in one of 11 listed areas, a qualified person must meaningfully review it with authority to change it, and the affected people must be told that AI was used. The areas are legal, medical, finance, credit, insurance, housing, employment, education and credentials, healthcare access, public benefits and services, and legal status and adjudication. The new text spells out what is excluded, including general education, wellness content and internal drafting.
The new requirements are for hardware. When Claude drives equipment that takes autonomous physical actions and could injure someone, a qualified operator must be able to observe the equipment and stop it. The equipment must hold a safe state if Claude is disconnected. The live policy text adds that operating limits must be enforced independently of the model's output. A robotics or industrial-control deployment that relies on the model itself to stay within safe bounds does not meet that standard.
The Supported Regions page also states its scope more precisely. The restriction covers people physically located in an unsupported region, companies incorporated or headquartered there, and entities majority-owned or controlled from there.
How the cruelty rule is enforced
Anthropic says Claude's ability to end these interactions "will remain the primary enforcement mechanism." That refers to a feature it introduced in August 2025, which lets Claude close a conversation as a last resort in rare, persistently abusive exchanges. An ended chat cannot receive new messages, but the user can start a new conversation or edit and retry earlier messages. The general enforcement section of the policy still says Anthropic may warn, throttle, limit, suspend or terminate access for violations. The October post does not say whether abuse of the model alone could lead to that.
A skeptical note: the clause is written broadly enough that how it is applied matters more than its wording. Anthropic has defined the excluded cases clearly. How often Claude ends a conversation under the new rule, and whether that ever escalates to account action, will only show up in practice.
Por que isso importa
- The rules that govern Claude use change today for every Anthropic customer, consumer and API alike.
- Teams deploying Claude in regulated decisions or physical hardware have concrete compliance requirements, including enforcing operating limits outside the model.
- Political and civic teams regain options that the old blanket ban on campaign targeting ruled out.
Pontos principais
- Anthropic's updated Usage Policy, announced October 8, is in force as of November 12, 2026.
- Deceptive-campaign rules are consolidated into one section, and the blanket ban on personalized vote and campaign targeting is gone.
- Weapons and surveillance language is tighter. Anthropic says the weapons changes reflect how it already enforced the policy.
- High-risk recommendations in 11 areas still require qualified human review and disclosure, and new rules cover hardware that takes physical actions.
- The ban on sustained, needless cruelty toward Claude is enforced mainly by Claude ending the conversation.
Fontes
- AnthropicPrimária2026 usage policy updateanthropic.com
- AnthropicPrimáriaUsage Policy (effective November 12, 2026)anthropic.com
- AnthropicPrimáriaClaude Opus 4 and 4.1 can now end a rare subset of conversationsanthropic.com