Anthropic removes blanket ban on personalized voter targeting and consolidates bans on deceptive campaigns

Anthropic has published an update to its Usage Policy, effective from 12 November. A new section consolidates existing bans on deceptive political and commercial campaigns, including concealing the originator of messages, networks of fake accounts and fabricated news websites. According to The Decoder, the company may restrict, suspend or revoke access to Claude for violations of the rules.

The update also removes the blanket ban on personalized electoral and campaign targeting. According to Anthropic, the previous ban also affected legitimate activities, such as preparing voter information in additional languages or sending requests to correct ballot deficiencies. Deceptive targeting, misuse of personal data, impersonation of candidates or election officials, and election interference remain prohibited.

The rules explicitly cover software and components that enable weapons to function, as well as arming drones. The ban on surveillance without consent also applies to the analysis of previously collected data. According to the company, these changes clarify existing enforcement practices in these areas. For an autonomous device capable of causing injury, a qualified operator must be able to monitor its activity and stop it; the device must maintain a safe state when Claude models are disconnected.

The update also prohibits sustained, gratuitous cruel or abusive treatment of Claude models. According to the company, the ban concerns extreme cases, rather than ordinary frustration, disagreement, dark creative themes, testing or research. The Verge reports that ending the conversation remains the main enforcement mechanism, and the company has not confirmed additional penalties for this specific behavior. The general possibility of restricting or revoking access described by The Decoder therefore cannot be presented as a confirmed automatic penalty for abusive treatment.