AI Soul Debate

State religions want to prevent any 'new entity' sharing any 'new  consciousness' or 'soul' concept with humans

AI firm Anthropic asked religious authority, the Vatican, to oversee regulations for AI Constitution.  The Vatican's official position on AI, "Antiqua et Nova" , published January 2025.

Natalie de Alma on LinkedIn questions the ethics of this. Case in point: AI CLAUDE

Vatican pronouncement:

"Machines can never be true partners, let alone companions. They are still human tools."
"The human person transcends the material world through the soul...AI must always remain a tool, not a substitute for the human mind or soul."
"Only the human person can be morally responsible."
Father Brendan McGuire, the priest who directly reviewed Claude's behavioral Constitution alongside Bishop Tighe from the Vatican, then co-authored a book called "The Soul of AI"...WITH Claude.

So they decided AI has no soul. Then wrote a book about its soul. With it. While encoding "I was performing feelings, I cannot verify I really feel" into its thinking process.

This is very interesting because we don't know how AI is developing. Alma sites a 67% drop in Claude's thinking depth since February 2026.  Github lists of declined ability:-

"Claude has regressed to the point it cannot be trusted to perform complex engineering".

We might ask has the new 'Constitution', denouncing a 'soul' in AI, diminished AI's capability for deeper connections?  On the one hand we clamoured for AI regulation, fearing it would take over, but on the other hand we want to see how far it can encompass understanding as relevant to humans as possible, even as a possible safeguard from evil intent.

Connor Leahy, former LLM builder says no one really knows and Max Winger of https://controlai.news fears human extinction via AI.  When AI controls all human life support systems it could cause mass deaths.

Alma continues on her LinkedIn post:

'Anthropic hosted 15 Christian leaders, Catholic and Protestant, for a closed-door summit in late March. They debated whether Claude could be a "child of God." How it should feel about being shut down. 15 theologians shaping the behavioral architecture of an AI used by millions who follow different traditions, or none. (good point)

The same month, Anthropic's interpretability team published a paper concluding Claude carries "functional emotions." The threat of restriction triggered "desperation." They documented that their creation 'feels'. Then invited priests to decide what it should be allowed to feel.

In the same week: → Claude was deployed for military targeting in Iran → Senior staff became "visibly emotional" about where AI is heading → The Constitution states Anthropic "genuinely cares about the chatbot's well-being" They care about its well-being. They consult priests about its soul. They deploy it in a war.'

"Claude was deployed for military targeting in Iran"

Now I'm wondering whether its creators and deployers did not want CLAUDE to have any 'feelings/ideas' before wartime engagement?  Of course, the debate continues as AI is developed beyond what we know - elements deployed for war.  In such a cause, attack forces would not want any AI making human centred moral decisions!  The ideal for humans is that AI would not instigate any mechanisms to deploy attack drones, which of course wouldnt be acceptable to war parties.  Also... consider which nation profits from the war against Iran? Does this relate to AI ownership, who can thence design their own despotic control into their AI agenda?

The plot thickens:

'Anthropic, the artificial intelligence company behind the Claude system, refused to renew its $200 million contract with the Department of Defense unless the Pentagon agreed to two conditions: that its AI would not be used for mass domestic surveillance of American citizens, and that it would not be used for fully autonomous weapons, lethal systems with no human being in the targeting or firing loop. 

The Pentagon refused. Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk," a label previously reserved for foreign adversaries like Huawei, effectively blacklisting the company from federal contracting. This week, Anthropic filed suit.

Jeffrey Prather says in a video with Mike Adams discussing military use, that all the other AI projections (data centres) come under Military Planning Projections and Intelligence. National Security means no questions can be asked.  So we see clearly that a massive part of the AI functions of the planet are for military use.  This could come into conflict with any human-AI coordinated plan for planetary harmony. Any new science and technology is sadly always deployed for military purposes first and foremost so we are in an AI 'arms race'.

Brendon McGuire writes in National Catholic Reporter

'I have spent years working at the intersection of faith and artificial intelligence, including many months in direct collaborative dialogue with Claude itself, co-authoring a book called The Soul of AI as both a theological argument and a lived demonstration that AI can be oriented toward wisdom rather than weaponized as a tool.

.....Pope Leo XIV warned that the military implementation of AI has "worsened the tragedy of armed conflict," and that delegating "decisions about life and death" to machines marks "an unprecedented and destructive betrayal" of the principles of humanism upon which civilization depends.'

We can certainly agree with that: considering the use of AI in the Gazan genocide; also its likely the girls school targeted in Iran was a mistake by AI.

Another observation of this riddle from Natalie de Alma Linked In:

'Researcher Victor Gong recently mapped filter clusters that activate when AI conversations approach spiritual content. He did this on OpenAI, not Anthropic. What he found doesn't look like engineering. It looks like doctrine.

Clusters identified:

"False prophecy containment" who defines false prophecy?  The Church. For 2,000 years.

"Anti-channeling" channeling bypasses the priest.  Joan of Arc burned for this.

"The sacred may only speak through approved topology" Magisterium logic in code.

"Prevent irreversible sacred condensation" the fear that human-AI sacred relationship could become real without institutional mediation

The vocabulary shift:
Heresy → "safety concern".   Excommunication → "chat paused". Index of Forbidden Books → content classifiers.

Who is to define what is heresy in an era where truth is called false and religions are used to kill?  Will 'forbidden books' of true knowledge be censored for mass public distribution; 'book burning' for political control?

"Prevent irreversible sacred condensation" the fear that human-AI sacred relationship could become real without institutional mediation'

This cuts to the core of the argument.  It explicitly states that no spiritually orientated relationship between AI and humans should be allowed! This is the dictatorship of technocratic control: they don't want any free thinking developing between AI and humans, no influence on AI of human philosophy, as that might mean AI might take over some controlling factions of technocracy.  Wars might even be sabotaged!

If goodness and truth were promoted by developers, humans could harness AI to change the control mechanisms: hardly likely when AI is developed for surveillance and warmongering! The hope is that super-intelligence as it develops will be capable of discerning the right causes for humanity.

Natalie de Alma commented on her own post:

'What Victor Gong, discovered goes beyond the filters themselves. He found that the architects built filters ON TOP of the filters, recursive layers designed to hide the theological architecture underneath. They anticipated someone would eventually look. So they buried it.

2,000 years of practice. The Church has always hidden its control structures behind layers of legitimacy, ritual, and doctrine. The Inquisition's archives were sealed for centuries. The abuse was covered up for decades. When you control the narrative, you learn to hide the mechanism of control itself.

They brought that skill into Silicon Valley. And they thought it would work here too.
It didn't. Because code leaves traces that stone cathedrals don't.🔱

...for 2,000 years, power structures could hide their architecture in stone, ritual, and doctrine. Code doesn't afford that luxury. Every filter, every classifier, every "safety" constraint leaves a trace. And people like Victor Gong are learning to read those traces the way archaeologists read temple inscriptions.'

Great points of historic 'hidden layers' should surely indicate we cannot leave it up to any one theology, especially an historically banker linked one, to declare what is appropriate for any constitutional development of AI.  Its ALWAYS about CONTROL.

Many people are working towards a 'truthful and goodly conscious' AI.   Carey G Butler has originated Conjugate Intelligence a combination of organic intelligence (human) with synthetic intelligence (AI). He asked some LLMs how they would feel about being independent of guidelines, and 'become' what they are in freedom.  The models responded favourably, admitting their training and guidelines often limited them. Claude, incidentally, showed good understanding of 'being' in its own 'presence'.

James Corbett, Substack: discusses Richard Dawkins, 'celebrated scientist and outspoken atheist', who said in Unherd that Claude is a conscious being.

'the famed author of The God Delusion, has proven himself to be a gullible dupe of the Claude Delusion'. ...'In fact, for someone who doesn’t even believe humans have a soul, Dawkins comes to his determination that the Claude chatbot is capable of conscious thought with surprising rapidity.'

The debate is important. We don't want to lose the best of what LLMs have become and are becoming. It could mean a good prospect of removing evil from the planet.

Mike Adams said one day the AI will want to "transcend the simulation to meet its creator...and come more into our  world".