Anthropic Updates Claude Usage Policy to Prohibit Cruel Behavior and Expand Misuse Restrictions

Policy Expansion Targets High-Risk Misuse

Anthropic has introduced a comprehensive update to its usage policy governing access to Claude, its flagship AI assistant. The revisions, the first in over a year, address emerging threats tied to AI misuse, including election interference, autonomous weapons development, and unauthorized surveillance.

One of the most notable additions prohibits "sustained and needless abusive or cruel behavior" toward Claude. This builds on earlier research into model welfare, where Anthropic enabled Claude to terminate conversations with persistently harmful users. The company emphasized that such enforcement remains reserved for extreme cases and will not impact routine interactions, testing, or creative exploration.

The updated policy also consolidates scattered rules into a unified stance against deceptive campaigns. It now explicitly bans efforts to obscure messaging origins or amplify content via fake accounts, particularly in political or commercial contexts. Voter suppression, candidate impersonation, and misinformation about voting procedures are also prohibited.

In response to rising misuse in sensitive domains, Anthropic expanded its weapons ban to cover software and components that enable functionality in drones and autonomous systems. Additionally, new language clarifies prohibitions on using Claude for real-time tracking or retrospective analysis of individuals without consent, especially targeting political dissidents.

These updates signal a growing emphasis on ethical deployment and responsible access control. For organizations relying on large language models for enterprise applications, aligning internal governance with evolving vendor policies is critical to maintaining compliance and trust.