Anthropic bans extreme model abuse and revises election rules

Where Anthropic draws the line on model abuse

Anthropic has updated its usage policy to prohibit sustained, needless abuse or cruelty toward its artificial intelligence models, according to reports from Engadget and The Verge. The update also revises rules governing election activity, Engadget reports.

For users, the key boundary is between ordinary disagreement and extreme abuse. Engadget reports that Anthropic still permits frustration, pushback and dark themes in creative work. Those stated exceptions leave room to challenge a model’s answers or explore difficult subjects.

The restriction follows an earlier change that allowed Claude to end conversations with persistently abusive users, according to Engadget. The new prohibition makes sustained, needless abuse a usage-policy issue, extending the approach beyond Claude’s ability to stop a conversation.

Election rules shift toward specific harmful conduct

Engadget reports that the revised election policy, titled “Do Not Undermine Democratic Processes,” focuses on falsehoods about candidates or voting procedures. It also addresses impersonation of candidates or election officials and turnout suppression—efforts to keep people from voting.

Anthropic also removed its blanket ban on personalized voter targeting, according to Engadget. The company’s stated reason was that the broad restriction could obstruct benign work, including translating voter guides or sending ballot cure notices.

The practical distinction is between removing a broad targeting ban and retaining restrictions on specific election harms. The reported change makes room for the helpful activities Anthropic identified; it does not establish that every form of personalized voter targeting is permitted.