First overhaul in more than a year: the ban on weapons extends to control software and armed drones, surveillance is better supervised and Claude's free harassment is outlawed.
Anthropic is changing its use policy for the first time in more than a year, to cover new high-risk abuse cases: election interference, weapons development, surveillance, medical and financial uses. The ban on weapons, already in place, is extended to software and components that make them work and to the arming of drones or autonomous vehicles, after several attempts to hijack Claude observed by the company.
Surveillance, propaganda and robots
Tracking people without their consent is now explicitly prohibited, both in real time and after the fact, as is the use of Claude to decide who should be investigated or arrested. The scattered rules on disinformation are grouped together in a ban on deceptive commercial or political campaigns (fake accounts, hidden messengers). When a pilot model of the material capable of harming, a qualified operator must be able to observe and stop it. However, some restrictions can be adapted in contracts with government clients.
The most commented change: the text prohibits "abusive or cruel, prolonged and gratuitous" behavior towards Claude. In line with its work on the "well-being of models", Anthropic specifies that the rule is aimed at extreme cases and not ordinary frustration, dark creative themes or tests. The main sanction remains the interruption of the conversation.
Read the full article on the original source
Go to the source →