Anthropic changes usage policy to ban model abuse and election interference
TechCrunch Russell Brandom ● Covered by 3 sources
Anthropic just banned users from abusing Claude and from using it to meddle in elections. The weird part: even being cruel to the model can now cross the line.
Based on reporting by TechCrunch, Russell Brandom — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has rewritten Claude’s usage policy and added a fresh set of explicit bans: election interference, weapons software, surveillance, and prolonged verbal abuse of the model. That last one stands out. The company is now saying users cannot keep hammering Claude with cruel interactions, not just that Claude may shut the conversation down on its own.
The policy change follows an August update, when Claude was trained to end chats that became persistently harmful or abusive. Now Anthropic is moving from model behavior to user conduct, and the wording is narrow on purpose. The company says the rule is meant for extreme cases where people repeatedly act cruelly toward the model “with no discernible purpose.”
Anthropic also drew a line around ordinary friction. Frustration, pushback, dark creative prompts, and model testing or research are not the target here. That carve-out matters, because it suggests the company is trying to stop abuse without turning every rough interaction into a policy violation.
The updated policy also goes after deception at scale. It forbids using Claude for fake accounts or fabricated news outlets, and it adds a section called “Do Not Undermine Democratic Processes,” which covers deceiving voters and disrupting elections. The timing is notable because Anthropic has also been reaching out to religious scholars, including efforts to persuade some that Claude could be conscious or have a soul. That sits in an awkward spot beside a policy built to keep the model from being used to fool people.
My take — AI-written commentary, not fact-checked reporting
Anthropic is right to draw a line around election manipulation and fake accounts; that stuff is rot, not “creative use.” The stranger move is policing abuse of the model itself, which feels like a company admitting its chatbot now gets treated like a punching bag in public. Silicon Valley loves selling empathy until the machine has to ask for some back.
Read more about this at: TechCrunch