---
title: Anthropic bans abuse of Claude in updated usage policy
url: https://www.elseif.net/anthropic-bans-abuse-of-claude-in-updated-usage-policy
published: 2026-10-11T21:05:29+00:00
language: en
section: Models
source: https://www.itmedia.co.jp/news/article/2610/09/2000002149/
organizations: Anthropic, Claude
publisher: elseif
---

# Anthropic bans abuse of Claude in updated usage policy

Anthropic announced on October 8 that it has revised the Usage Policy for its AI model Claude, effective November 12. The update clarifies existing rules and adds new provisions targeting abusive behavior. The company cited increased autonomy of Claude over the past year as the reason for expanding use cases. A new rule prohibits continuous and unnecessary cruel or abusive treatment of AI models, though it applies only to extreme cases without clear purpose, excluding general complaints, rebuttals, dark creative themes, and model testing or research. Anthropic already enabled Claude and Claude Code to terminate conversations with abusive users, and the new rule aligns with this capability, which will remain the primary enforcement method.

A new section consolidates prohibitions on deceptive campaigns and artificial activity, covering state media, propaganda outlets, and private firms using Claude to run fake accounts or fabricated news sites. Previously scattered across election, fraud, privacy, and misinformation rules, these are now grouped under Do Not Engage in Deceptive Campaigns or Artificial Activity, applying regardless of political or commercial intent. The policy also bans hidden source disclosure, artificial account amplification, and infrastructure building for influence operations. Election-related rules now specifically prohibit voter deception or election interference. Weaponization and surveillance rules were clarified: arming software, components, or drones is prohibited, and unauthorized tracking or use in law enforcement decisions is banned.

Anthropic stated these do not change enforcement but formalize prior practices. For high‑risk applications in health, legal rights, finance, and similar domains, the policy requires qualified human oversight of AI suggestions and disclosure to affected individuals. This requirement remains unchanged, but the company now explicitly lists applicable and exempt activities across eleven sectors, including physical equipment control under the Model Hardware Standard introduced in August. Requirements include human monitoring and shutdown capability, and ensuring safe state retention if Claude connection drops.