The new usage policy targets users who mistreat Claude for no apparent reason, and includes a section dedicated to safeguarding democratic processes.
On October 8, 2026, Anthropic updated its usage policy. The most commented novelty concerns prolonged verbal abuse of its models: the company specifies that the rule only applies in extreme cases, when a user is repeatedly cruel and without an identifiable purpose. Ordinary frustration, contestation, dark themes in creative writing or testing and research on models remain explicitly out of scope. Since August, Claude has already been able to put an end to a persistent abusive conversation.
Elections, surveillance and weapons
A section entitled "Do not undermine democratic processes" prohibits voter deception, disruption of polls and large-scale disinformation campaigns, such as fake accounts or fabricated media. The policy also mentions the development of weapons software and surveillance practices.
This update comes as Anthropic recently spoke with theologians about the question of Claude's possible conscience, a subject on which the company is cautious in principle.
Read the full article on the original source
Go to the source →