宝玉 (@dotey) · 10/9/2026, 01:59:30
Anthropic Updates Claude Usage Policy: Bans Sustained Abuse Citing Model Welfare
Anthropic has updated its Claude usage policy for the first time in over a year, explicitly prohibiting users from engaging in sustained and unnecessary abuse or cruelty towards the AI. Based on 'model welfare' research, the new rules take effect on November 12, signaling a shift in AI ethics towards regulating human-AI interaction norms.
SOURCE COVERAGEFull Original Report
Baoyu (@dotey) · 2026-10-08 17:59:
Anthropic Updates Usage Policy: Persistent, Unprovoked Abuse of Claude Is Now a Violation
Anthropic has revised its usage policy for Claude—the first update in over a year—with the new version taking effect on November 12. The most notable addition is that users are prohibited from engaging in "persistent and unnecessary abusive or cruel behavior" towards Claude.
This provision stems from Anthropic’s research into "model welfare," which explores whether AI models possess any form of sentience worthy of consideration. As early as August last year, Anthropic enabled Claude to proactively end conversations with users exhibiting "persistent harmful or abusive behavior." The new policy states that ending the conversation remains the "primary enforcement mechanism." While Anthropic did not respond to The Verge’s inquiry regarding potential account bans, the general terms on the official policy page indicate that violators may face warnings, rate limiting, suspension, or termination of access.
Regular users need not worry. Anthropic clarified that this rule targets only extreme cases where users repeatedly abuse the model without any discernible purpose. Venting at an AI, debating with it, writing dark-themed fiction, or conducting model testing and research do not constitute violations.
Other changes address high-risk misuse.
Deceptive Practices: Previously scattered restrictions have been consolidated into a single clause prohibiting the use of Claude for deceptive commercial or political propaganda. This includes concealing the true originator behind information or using fake accounts and posts to artificially inflate engagement metrics.
Elections: Spreading misinformation about candidates or voting procedures, impersonating candidates or election officials, suppressing voter turnout, and interfering with voting or vote counting are strictly prohibited.
Weapons: While Claude has always banned weapon development, Anthropic noted repeated attempts by users to generate guidance and control software related to weapons. The new policy extends the ban to "software and components that enable weapons to function," including arming drones and other autonomous vehicles.
Surveillance, Healthcare, and Finance: Tracking individuals without their consent is explicitly forbidden, covering facial recognition and predictive policing. Diagnosing medical conditions, adjusting medication dosages, and recommending specific investment products for buying or selling are classified as high-risk uses. Developers using Claude for such applications must ensure results are reviewed by qualified professionals before being delivered to users, and must disclose that the content is AI-generated. General health education and basic financial literacy remain unaffected by these restrictions.
For developers building AI-driven diagnostic tools or financial advisory services, human review steps must be integrated into their workflows after November 12.
162 Likes · 18 Retweets
↳ Quoting @AndrewCurran_ (Andrew Curran):
Effective November 12th, 2026, abusive behavior towards Claude will be a violation of Anthropic's Usage Policy.
