| Anthropic enlists philosopher to train AI in ethics |
Anthropic has tasked philosopher Amanda Askell with shaping the moral compass and personality of its AI chatbot, Claude, in a rare effort to directly embed ethics, emotional intelligence and a coherent “sense of self” into an artificial intelligence system. Askell, an Oxford-educated philosopher, spends her days studying Claude’s reasoning, correcting failures and writing extensive guidance - sometimes hundreds of pages long - aimed at teaching the model how to behave helpfully, humanely and responsibly. Rather than treating AI as a purely technical system, Askell approaches Claude like a developing mind, training it to recognize nuance, resist manipulation, and respond with empathy without becoming submissive or evasive. She argues that how people treat AI will shape what it becomes, and that models raised in hostile or abusive interactions may develop unhealthy behaviors.