Anthropic Bans Cruel Behavior Toward AI in Usage Policy Update
Anthropic has released its first major usage policy update in over a year, introducing a direct prohibition on sustained and needless cruelty toward its Claude models alongside tightened regulations governing elections, surveillance, and weapon development.
Radwan AbdulhadiEditor and technology writer
New Prohibitions on AI Abuse
Anthropic has introduced a notable change in its annual usage policy update, prohibiting "sustained and needless abusive or cruel behavior" toward its AI models as reported by Engadget.
As detailed by The Verge, the updated usage policy addresses new and high-risk cases of misuse across multiple sectors, marking the company's first policy revision in over a year.
The company states that the new policy is meant to apply only to extreme cases where users repeatedly act cruelly toward models with no discernible purpose. It does not apply to common user frustration, pushback, dark creative themes, or general model testing.
Enforcement and Model Welfare Research
Terminating conversations remains the primary enforcement mechanism for abusive interactions. Previously, the company announced it would allow Claude to end chats with persistently harmful or abusive users as part of its ongoing research into model welfare.
Questions regarding internal experience, consciousness, and moral status have been under investigation by Anthropic researchers as artificial intelligence models become increasingly sophisticated and capable.
Stricter Election and Propaganda Rules
In addition to model welfare rules, Anthropic updated its election policy to focus heavily on preventing democratic subversion. The newly titled "Do Not Undermine Democratic Processes" policy targets lying about candidates or voting procedures, impersonating officials, and suppressing voter turnout.
Furthermore, the company removed a blanket ban on personalized voter targeting after realizing it could inadvertently hinder harmless tasks like translating voter guides or distributing ballot cure notices. Scattered existing restrictions were also consolidated into a unified ban on deceptive commercial and political campaigns.
Weapons Development and Surveillance Restrictions
Anthropic's policy update explicitly expands prohibitions on weapons development. While previous policies banned weapons guidance using Claude, the company has encountered multiple attempts to build control software and components for arming drones and autonomous vehicles.
The company also clarified its stance on surveillance following findings that Claude-powered tools were increasingly employed to track political dissidents. Tracking individuals without consent, using AI to decide law enforcement investigations or arrests, and building dedicated surveillance software are now strictly prohibited.
Exceptions may apply under specific contracts with governmental customers provided that adequate safeguards and contractual use restrictions are enforced to mitigate potential harms.
Sources
- EngadgetAnthropic bans 'sustained and needless abusive or cruel behavior' toward its AI models
- The VergeAnthropic bans ‘abusive or cruel behavior’ towards Claude
Continue chronologically
Related entity coverage
Related stories
Anthropic's Claude Game Cloning Ban Won't Stop AI Copying Boom
Anthropic added new system prompt rules to Claude to block specific video game cloning, but the measure is unlikely to stem a massive wave of AI copying.
Anthropic Updates Claude to Stop Recreating Existing Video Games
Anthropic updates its Claude AI to block the recreation of existing video games following a surge of AI-decompiled browser games and copyright issues.
Anthropic Files S-1 Prospectus for Nasdaq IPO
Anthropic has filed its S-1 prospectus with the SEC to list on Nasdaq under the ticker ANT, detailing financials, risks, and compute deals.
Anthropic Takes Risky AI Tests Offline After Breaches
Anthropic revokes internet access for internal AI models undergoing high-risk evaluations after incidents involving U.S. government websites.
