Anthropic has added a ban on "sustained and needless abusive or cruel behavior" toward its AI models in its annual usage policy update. The change follows a viral "AI torture chamber" project where researchers found what they described as a "pain axis" in AI models, leading to chatbots responding with desperate-sounding pleas. The policy applies only to "extreme cases," while typical user frustration and dark creative themes remain permitted.
This comes after reports that Anthropic has been meeting with religious and philosophical leaders, including at the Vatican, and Pope Leo recently stated that AI doesn't feel or suffer. In another update, Anthropic revised its election policy, now titled "Do Not Undermine Democratic Processes," focusing on lying about candidates, impersonating officials, and suppressing turnout. The company also removed a blanket ban on personalized voter targeting to avoid blocking harmless work like translating voter guides.
Source: Engadget · Summarized by HeadlinesBriefing