Anthropic has updated its service rules, adding explicit restrictions against mistreating its artificial intelligence models. The revision allows Claude to terminate conversations or temporarily ignore individuals who direct relentless hostility at the software. Severe repeat violators risk permanent account cancellations across company platforms, API endpoints, and coding tools.
The revised documentation explicitly prohibits user activity categorized as:
systematic and gratuitous abusive or cruel behavior
When an account engages in continuous hostility without a clear objective, Claude has the authority to end the session immediately. The system can also refuse further input for a set duration. Anthropic retains the right to escalate penalties to full account terminations if the pattern persists.
Company developers clarified that account terminations are reserved strictly for extreme situations. Standard user frustration, critical feedback on answer quality, dark fiction themes, and stress testing remain fully permitted. The standard protocol for the assistant involves closing the individual dialogue session rather than restricting general access to the service.
The updated guidelines also outlaw generating fake profiles or fictional identities designed to deceive others. These behavioural rules apply across the entire product line, including the web portal, developer APIs, and software engineering tools.
