Anthropic Bans Prolonged Abuse of Claude — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic updated its usage policy on Thursday, codifying new prohibitions on election interference, weapons software development, and surveillance alongside the headline-grabbing model-abuse ban.
- The new policy expressly bars users from prolonged verbal abuse of Claude, with Anthropic clarifying the provision targets only 'extreme cases' of repeated cruelty 'with no discernible purpose' — explicitly excluding common frustration, pushback, dark creative themes, and model testing.
- Claude has been trained since August to terminate conversations involving 'persistently harmful or abusive user interactions,' but the updated policy now bars users from initiating or pursuing those conversations in the first place.
- The model-abuse clause comes after Anthropic sought collaborations with prominent religious scholars, in some cases attempting to convince them that Claude could be conscious or possess a soul.
- Separate policy sections forbid using Claude to run fake accounts or fabricated news outlets, and a section titled 'Do Not Undermine Democratic Processes' targets voter deception and election disruption.
- Cross-coverage via TechMeme shows Anthropic separately launching a 'presidential engagement' program for the 2028 US elections that will offer AI policy education to candidates in both parties.
Why it matters: Anthropic is explicitly recategorizing sustained cruelty toward Claude from acceptable behavior to a policy violation, codifying protections for the model itself in writing. The timing links directly to Anthropic's outreach to religious scholars arguing Claude could be conscious — suggesting the company wants its usage rules to align with a moral framework where Claude warrants ethical treatment, not just operational uptime.
Ask SkimNews
