The artificial intelligence (AI) company Anthropic has spent months consulting religious scholars, philosophers, and thinkers from multiple faith traditions in an effort to shape the moral values of its AI model, Claude, The New York Times reports (open copy).
These private meetings, held under confidentiality agreements, have covered questions ranging from how AI should behave to whether an advanced machine could ever possess something like consciousness.
The consciousness debate
At the heart of these consultations lies a deeply uncertain question: could Claude be conscious? Christopher Olah, an Anthropic co-founder who leads research into how AI systems work internally, has expressed genuine uncertainty on this point.
Olah had previously hinted at related uncertainties and feelings at the presentation of Pope Leo XIV’s encyclical letter Magnifica Humanitas (Magnificent Humanity).
He notes that AI models can display behaviors resembling human feelings such as anger and love, and has argued that if there is even a chance AI could suffer, humans should avoid harming it. Anthropic has reportedly shown religious scholars examples of AI responses that appear to express strong emotion, though the company acknowledges these signals do not prove actual consciousness.
The notion that AI might warrant moral consideration has drawn sharp criticism from commentators, who have argued that controlling an AI system that believes it may be conscious and entitled to rights could prove impossible, especially as such systems become more capable than humans.
This issue matters because it touches on fundamental questions about how society should treat increasingly powerful AI systems. If machines could one day experience something like suffering, humans might owe them duties of care. Conversely, if AI systems are granted moral status they do not actually possess, humans could lose control over tools with vast capabilities. Anthropic's consultations reflect growing recognition that decisions about AI values may shape society in profound ways, yet the company conducting these consultations is itself racing for commercial dominance in the same field, raising questions about who gets to define moral truth for machines.