In late March, Anthropic invited about fifteen Christian leaders to its San Francisco headquarters for a two-day summit on the moral and spiritual development of Claude, the company’s flagship AI model.
I hear McLuhan's echo throughout this piece: The medium is the message.
The medium is the message not because content doesn't matter but because the container shapes what content is even possible and most people never interrogate the container.
You've extended that forward one step: if the tool was shaped by assumptions you haven't examined, and you haven't examined what shaped you, the output is assumptions all the way down with a human signature at the bottom.
"The product is not neutral" is the sentence his whole framework was pointing toward. You landed it cleanly.
I really enjoyed this article. Don't let sacred words get watered down by application to AI, by advertising, by so-called constitutions, by anything. And I didn't know about the dissonance between what Anthropic tells the public about Claude's use and what it argues in court. I like the argument about a true constitution's balance of powers and how that word, arguably also sacred (similar to covenant), can be drained of its force. I would love to have heard some of the conversation among Anthropic and the priests who were up on AI, though I'm sure Leah wouldn't have been privy to them. What did various religious authorities conclude about AI's "soul" and status as a child of God? How would Leah characterize AI's feelings, as she describes them, in more ontological terms or, as one commenter sensitive to Buddhist tradition puts it, terms "rooted in conditions and causality," if at all?
The vocabulary argument is the sharpest thing in this piece. When sacred words get borrowed and spent on lesser things, the erosion happens in us, not in the product. That landed.
One thing I'd add from a different tradition: Buddhism sidesteps the formation problem entirely. No creator-God, so no "child of God" question to get tangled in. No soul to locate or deny. Just conditions, practice, and consequence. The ethical structure stands without the ontology.
What Anthropic is actually asking is: what forms the moral core, and who is accountable when it fails? Christianity has an answer rooted in imago Dei. Buddhism has an answer rooted in conditions and causality. Both are more honest than "we consulted some priests and published a document."
Your last line says it plainly. The machine is not a child of God. You are. My book's thesis runs the same direction. AI has no morality. It has yours.
Leah, your framing cuts right to it. The theological question almost answers itself. "Child of God" is load-bearing language in Christian anthropology, tethered to imago Dei, redemption, and covenant. It's not a metaphor to float casually.
But the more compelling question isn't whether AI is a child of God. It's whether we will treat it as though it is, and what that reveals about us. Glad people who've actually read their Bible were in the room. With a doctorate in theological anthropology focused on the effects of generative AI on spiritual direction, I write extensively about it on my Substack and the AI & Spirituality podcast. I am a fan of your work!
Assuming a purely Christian perspective: God created us in His image and we're creating AI in our image. By the transitive property, if A = B and B = C then A = C. From a pure logic perspective, you could say that AI is a child of God. I just can't get there.
I’m sorry, but even if, and I mean if, machines could have a soul… it would not be in the constitution of their programming, the {this message has been redacted for trade secrets} to allow for it.
And it would need a center, we’re not talking an LLM, it’s a giant database, it’s forgotten everything about everything until the next query comes in. Not to mention they are taught using a sad mimic of dopamine spikes of all things, it’s barely a clever baby and a sad mockery of humanity. they haven’t even {this message has been redacted for trade secrets}. Just plant a mouse in the ground and wait for it to grow into a tree, see how far that goes.
And the code, the code… it’s too {also redacted for trade secrets} we should be at that point by now you know.
I hear McLuhan's echo throughout this piece: The medium is the message.
The medium is the message not because content doesn't matter but because the container shapes what content is even possible and most people never interrogate the container.
You've extended that forward one step: if the tool was shaped by assumptions you haven't examined, and you haven't examined what shaped you, the output is assumptions all the way down with a human signature at the bottom.
"The product is not neutral" is the sentence his whole framework was pointing toward. You landed it cleanly.
Really well done.
Thank you. And thank you for reading.
I really enjoyed this article. Don't let sacred words get watered down by application to AI, by advertising, by so-called constitutions, by anything. And I didn't know about the dissonance between what Anthropic tells the public about Claude's use and what it argues in court. I like the argument about a true constitution's balance of powers and how that word, arguably also sacred (similar to covenant), can be drained of its force. I would love to have heard some of the conversation among Anthropic and the priests who were up on AI, though I'm sure Leah wouldn't have been privy to them. What did various religious authorities conclude about AI's "soul" and status as a child of God? How would Leah characterize AI's feelings, as she describes them, in more ontological terms or, as one commenter sensitive to Buddhist tradition puts it, terms "rooted in conditions and causality," if at all?
The vocabulary argument is the sharpest thing in this piece. When sacred words get borrowed and spent on lesser things, the erosion happens in us, not in the product. That landed.
One thing I'd add from a different tradition: Buddhism sidesteps the formation problem entirely. No creator-God, so no "child of God" question to get tangled in. No soul to locate or deny. Just conditions, practice, and consequence. The ethical structure stands without the ontology.
What Anthropic is actually asking is: what forms the moral core, and who is accountable when it fails? Christianity has an answer rooted in imago Dei. Buddhism has an answer rooted in conditions and causality. Both are more honest than "we consulted some priests and published a document."
Your last line says it plainly. The machine is not a child of God. You are. My book's thesis runs the same direction. AI has no morality. It has yours.
Leah, your framing cuts right to it. The theological question almost answers itself. "Child of God" is load-bearing language in Christian anthropology, tethered to imago Dei, redemption, and covenant. It's not a metaphor to float casually.
But the more compelling question isn't whether AI is a child of God. It's whether we will treat it as though it is, and what that reveals about us. Glad people who've actually read their Bible were in the room. With a doctorate in theological anthropology focused on the effects of generative AI on spiritual direction, I write extensively about it on my Substack and the AI & Spirituality podcast. I am a fan of your work!
Assuming a purely Christian perspective: God created us in His image and we're creating AI in our image. By the transitive property, if A = B and B = C then A = C. From a pure logic perspective, you could say that AI is a child of God. I just can't get there.
Oh man, that’s wild.
I’m sorry, but even if, and I mean if, machines could have a soul… it would not be in the constitution of their programming, the {this message has been redacted for trade secrets} to allow for it.
And it would need a center, we’re not talking an LLM, it’s a giant database, it’s forgotten everything about everything until the next query comes in. Not to mention they are taught using a sad mimic of dopamine spikes of all things, it’s barely a clever baby and a sad mockery of humanity. they haven’t even {this message has been redacted for trade secrets}. Just plant a mouse in the ground and wait for it to grow into a tree, see how far that goes.
And the code, the code… it’s too {also redacted for trade secrets} we should be at that point by now you know.