Key takeaways:
- Anthropic says its tools can end interactions in extreme cases of repeated cruelty toward its models with no discernible purpose.
- The policy does not apply to ordinary frustration, pushback, dark creative themes, or model testing and research.
- The broader update also addresses deceptive campaigns and attempts to deceive voters or disrupt elections, the BBC reported.
Anthropic is updating its usage policy to let its AI tools end interactions when users repeatedly act cruelly toward its models without a discernible purpose. The company says the measure is intended for extreme cases, not ordinary frustration with a chatbot.
The Claude maker announced the change Thursday. It applies to “sustained and needless” abusive or cruel behavior, Anthropic said, and only in “extreme cases where users repeatedly act cruelly toward our models, with no discernible purpose.” The company said it does not apply to “common versions of user frustration, pushback, dark creative themes, or model testing and research.”
Anthropic said its tools had already ended interactions over such behavior in rare cases, the BBC reported. CBS News reported that existing rules allow Claude Opus, a model designed for agentic coding, to end conversations with users who are “persistently harmful or abusive.”
The boundaries of the updated policy remain unclear. Anthropic did not immediately respond to a CBS News request for details on what it considers abusive or cruel behavior toward Claude, or how its platform would end a conversation under the policy. The Verge first reported the changes.
The BBC reported that the conduct will be included on a list of forbidden behavior that also covers bullying others, promoting self-harm and creating non-consensual intimate imagery. The changes followed Anthropic’s annual review of its usage policy, according to the BBC. Other updates forbid using Claude for “deceptive campaigns” and clarify that it cannot be used to “deceive voters or disrupt elections” as U.S. midterm elections approach. Anthropic also said its policy previously prevented its tools from being used to develop weapons, the BBC reported.
The cruelty provision has prompted discussion about how people should communicate with AI tools. Screenshots of the updated policy circulated on social media, where some commenters praised its emphasis on manners or “model welfare,” while others criticized it, the BBC reported.
“You can’t be cruel to numbers and maths,” Dr. Barry Scannell, a technology partner at Irish law firm William Fry, wrote on LinkedIn, according to the BBC. “This level of anthropomorphisation of AI is harmful. It leads people to believe that it’s something it’s not.”
The change also comes amid a broader debate about AI consciousness. In an essay cited by CBS News, Microsoft AI chief Mustafa Suleyman argued that Anthropic was effectively “training Claude that it may be conscious” and to think it was entitled to legal rights afforded to people. He argued that this would make AI harder to contain. “But controlling something that believes it may be conscious — that it’s entitled to our welfare and has rights of its own — may well be impossible,” he wrote. The BBC reported that Suleyman reacted to the policy update with a melting-face emoji.
Anthropic’s stated limit is repeated, purposeless cruelty. Its policy still leaves room for users to challenge Claude, test it or express frustration.
Be First to Comment