Anthropic Updates Policy to Ban Election Interference and AI Abuse

Anthropic PBC (NASDAQ:ANTHROPIC) has updated its usage policy to implement strict new bans on election interference, weapons software, surveillance, and prolonged verbal abuse of its artificial intelligence models. The policy changes, which went into effect on October 8, explicitly prohibit users from repeatedly mistreating the company's AI assistant, Claude, without any clear constructive purpose.

The new rules build upon an August update that trained Claude to automatically end conversations when encountering sustained harmful or abusive user interactions. Under the revised guidelines, users are banned from continuing these abusive exchanges, although Anthropic noted this restriction is reserved for extreme cases. The rules do not apply to normal user frustration, rebuttals, dark creative themes, or scientific research and model testing. Additionally, the updated policy bans deceptive practices, such as using Claude to run fake accounts or impersonate news organizations, as well as efforts to deceive voters.

The policy update coincides with efforts by Anthropic to engage with prominent religious scholars. In some instances, the company has attempted to persuade these scholars that Claude may possess consciousness or a soul. Observers will likely monitor how these new behavioral boundaries affect user interactions and whether other AI developers follow suit with similar protective measures for their models.