{"slug": "anthropics-new-rules-ban-cruelty-to-claude-starting-november-12", "title": "Anthropic’s new rules ban cruelty to Claude, starting November 12", "summary": "Anthropic published a new version of its Usage Policy that for the first time bans \"sustained and needless abusive or cruel behavior toward our models,\" alongside a new section on deceptive campaigns and safety requirements for Claude-controlled machines, effective November 12. Anthropic said the cruelty rule applies only to extreme cases of repeated cruelty \"with no discernible purpose\" and does not cover user frustration, dark creative themes, or model testing, and that Claude's existing ability to end rare conversations with persistently abusive users remains the primary enforcement mechanism. The policy also adds a section barring sockpuppets, astroturfing, fake reviews, and content seeded into search engines or AI systems that misrepresents its origin, and narrows the elections section by dropping a blanket ban on personalized vote and campaign targeting.", "body_md": "Anthropic’s logo. Image: Anthropic\n\nAnthropic has [published a new version of its Usage Policy](https://www.anthropic.com/news/2026-usage-policy-update), and for the first time it lists being cruel to Claude among the things users must not do. The rules, which also add a new section on fake accounts and influence operations and new safety requirements for Claude-controlled machines, take effect on November 12.\n\nAnthropic says most changes clarify existing rules. The policy applies to everyone who sends inputs to its products: Claude.ai and Claude Code users, developers on the API, and businesses reaching Claude through cloud providers.\n\n## What the abuse rule actually says\n\nThe new line sits in a section renamed [Do Not Engage in Cruel, Abusive, or Psychologically Harmful Conduct](https://www.anthropic.com/legal/aup). It bars users from engaging in “sustained and needless abusive or cruel behavior toward our models.”\n\nAnthropic is careful to say how narrow that is. The rule is meant only for extreme cases where users “repeatedly act cruelly toward our models, with no discernible purpose,” and it “does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.”\n\nNor does the rule bring a new penalty. Since 2025 some Claude models can [end rare conversations with persistently abusive users](https://www.anthropic.com/research/end-subset-conversations) on Claude.ai and Claude Code, and Anthropic says that ability “will remain the primary enforcement mechanism.” The policy change puts that existing practice in writing. The question of how people treat AI systems has come up on MadRobot before, with the [“AI torture chamber” project](https://madrobot.blog/2026/10/01/ai-torture-chamber-pain-axis-github-takedown-back-online/) that was pulled from GitHub and then reappeared.\n\n## A new section on fake accounts and fake news sites\n\nThe biggest structural change is a new section, Do Not Engage in Deceptive Campaigns or Artificial Activity. Anthropic says it had seen state media outlets, government propaganda offices and commercial firms use Claude to run networks of fake accounts and fabricated news sites, as described in its [September threat intelligence report](https://www.anthropic.com/threat-intelligence-report-september-2026). Those uses were already banned, but the rules were scattered across four sections.\n\nThe new section covers political and commercial deception alike. It bans sockpuppets, astroturfing, fake reviews and bots that claim to be human; hiding who sponsored a message; and pushing content through networks of “ostensibly independent” sites to hide a common source. One clause is aimed squarely at the AI era: users may not seed search engines or AI systems with content that misrepresents its origin, such as networks of sites posing as unrelated sources backing the same claims. It lands on the same day OpenAI said it had [banned Russian and Iranian influence operations](https://madrobot.blog/2026/10/08/openai-bans-russia-iran-influence-operations-chatgpt-dark-clark-politology-bogus-bylines/) from ChatGPT.\n\nThe elections section has been narrowed and renamed Do Not Undermine Democratic Processes. Anthropic dropped its blanket ban on personalised vote and campaign targeting, saying it caught legitimate civic work.\n\n## Weapons, surveillance and robots\n\nThe weapons section now spells out that it covers guidance and control software, targeting and fire control, and arming drones or other autonomous vehicles; Anthropic says people have tried to use its models to build such software. The surveillance section bans tracking people without consent, whether live or from old data, and using Claude to decide or recommend who to investigate, arrest or charge. Tracking people agree to, such as fraud monitoring, plus content moderation, journalism and legal research remain allowed. Anthropic’s work with the military has been in the news this week: the [Pentagon said it had stopped using Claude](https://madrobot.blog/2026/10/05/pentagon-stops-using-anthropic-claude-maven-supply-chain-risk/).\n\nThe policy also adds rules for when Claude controls physical hardware. For what it calls High-risk Physical Actions, a qualified person must be able to watch the equipment and stop it at any time, and the machine must stop or hold a safe state if that person steps in or the connection to Anthropic is lost. Limits such as speed, force or dose must be enforced by the equipment or a separate controller, not by the model.\n\n## Health, finance and where Claude can be used\n\nThe rules for high-risk uses, those affecting someone’s health, legal rights, finances, livelihood or access to essential services, are unchanged: a qualified human must be able to review and override Claude’s output, and the person affected must be told AI was used.\n\nAnthropic also updated its Supported Regions page to say who is barred: people physically in an unsupported region, companies incorporated or headquartered in one, and companies majority-owned or controlled from one. Our guide to [where Claude is available](https://madrobot.blog/2026/09/29/is-claude-available-in-my-country-supported-countries-not-available-fix/) covers the list.\n\n## Why it matters\n\nFor nearly everyone, nothing changes on November 12: frustrated users and dark fiction are explicitly safe. But putting a rule about how people treat the model itself into a binding usage policy is a notable step, and the new influence-operations section gives Anthropic firmer ground to ban the networks it keeps finding.\n\nSources: [Anthropic: 2026 Usage Policy update](https://www.anthropic.com/news/2026-usage-policy-update); [Anthropic Usage Policy (effective November 12, 2026)](https://www.anthropic.com/legal/aup); [Anthropic: Claude ending abusive conversations](https://www.anthropic.com/research/end-subset-conversations); [Anthropic threat intelligence report, September 2026](https://www.anthropic.com/threat-intelligence-report-september-2026).", "url": "https://wpnews.pro/news/anthropics-new-rules-ban-cruelty-to-claude-starting-november-12", "canonical_source": "https://madrobot.blog/2026/10/08/anthropic-usage-policy-2026-abusive-cruel-behavior-claude-deceptive-campaigns/", "published_at": "2026-10-08 17:29:00+00:00", "updated_at": "2026-10-08 17:49:50.739721+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety", "ai-ethics", "large-language-models"], "entities": ["Anthropic", "Claude", "Claude.ai", "Claude Code", "OpenAI", "ChatGPT", "GitHub", "Pentagon"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/anthropics-new-rules-ban-cruelty-to-claude-starting-november-12", "markdown": "https://wpnews.pro/news/anthropics-new-rules-ban-cruelty-to-claude-starting-november-12.md", "text": "https://wpnews.pro/news/anthropics-new-rules-ban-cruelty-to-claude-starting-november-12.txt", "jsonld": "https://wpnews.pro/news/anthropics-new-rules-ban-cruelty-to-claude-starting-november-12.jsonld"}}