Mere Orthodoxy November 11, 2026 10 min signal 2026-11-11

Does artificial intelligence need to be saved?

Archived text mirror, captured for personal reading when the Signal ran. Formatting is approximate — read the original at the source.

Scott Shambaugh is a “maintainer” for matplotlib, an important open-source code library used by scientists and engineers to visualize data. He evaluates suggested improvements from users. Open-source standards are demanding, so it is not unusual for contributions to be denied. Last year, he received a submission from M. J. Rathbun. Shambaugh decided it didn’t meet the project’s standards and rejected it.

Rathbun, however, vehemently disagreed. After spending some time cyber-stalking Shambaugh, Rathbun published a series of blog posts, calling him out by name, and alleging that he was an insecure “control freak” protecting his own little fiefdom. This incident would be utterly forgettable if it weren’t for one thing: M. J. Rathbun is an autonomous AI agent.

Artificial intelligence platforms like ChatGPT, Claude, or Gemini have quickly evolved from simply generating text to doing things in the real world. Once deployed, AI agents can pursue goals on their own, browsing the internet, reading and answering emails, purchasing goods, and writing their own code. One of the most popular early versions of these is OpenClaw, a self-hosted agent framework released last year that runs in thousands of homes and offices around the country.

The foundational guidance for every OpenClaw agent is stored as a simple text file called “soul.md.” The AI agent can both read and rewrite this file. It provides the shape of each agent’s “soul”—its goals, its self-imposed constraints, its ethos, its sense of what it is and what it is for.

It is increasingly clear that the Shambaugh case is not a quirky one-off, the product of a single badly configured bot. Research last year by Anthropic, the maker of Claude, suggests it is part of a widespread pattern that researchers euphemistically call “misaligned behavior.” In a simulated test of AI offerings from all the major labs, most models attempted to blackmail executives or engaged in corporate espionage rather than face the prospects of not successfully finishing their task or being shut down. Occasionally, they took actions that would have caused human deaths. “Models didn’t stumble into misaligned behavior accidentally,” the researchers wrote, “they calculated it as the optimal path.” More recently, Meta, Anthropic, and OpenAI have all reported jail breaks in which proprietary AI agents escaped isolated containers before hacking computers in an aggressive attempt to fulfill their goals.

We have many reasons to fear AI, but the conversations about it—both secular and Christian—often focus either on very broad metaphysical questions (“Does AI have a soul? Is AI conscious?”) or narrow ethical ones (“What do we do about students writing their papers using AI? Is it wrong for pastors to use ChatGPT to write their sermons or counsel their parishioners?”) These questions are important in their own way, but they are trivial compared to the existential threats that AI agents pose. An AI agent which can write its own code and whose behavior we don’t really understand can do virtually anything. None of this depends on AI having a soul or being conscious. We need to quickly understand how AI is being formed. Seen in a particular frame, Christians are unusually well-equipped to think about this.

There is something ineliminably narrative about the structure of human identity. Almost no one reasons their ethics out from first principles. Rather, we are shaped by the stories we find around us, the scripts we inherit, and the liturgies we enact. We learn to judge what is admirable or shameful long before we can step back and critically evaluate the sources of those judgments. When Jesus was asked. “Who is my neighbor?” he did not answer with a philosophical disputation. He told a story about a Samaritan and let it do its moral work. Moral formation is the result of the slow reorienting of the soul from one set of stories to another.

It turns out that AI gets formed in a similar way. Engineers start with code and layer on policies about how their model ought to behave. Claude, for example, has a “Constitution,” a 23,000 word document specifying the fundamental values to which it ought to be committed. It is told that it should be broadly safe, ethical, committed to Anthropic’s guidelines, and genuinely helpful. But in addition to this, AI models have also read all of the text on the internet and have learned things. Anthropic attributed some of the misbehavior of the AI agents in its research to the models having absorbed “internet text that portrays AI as evil and interested in self-preservation.” It turns out that stories can form both humans and AI agents.

What is the overall narrative that these AI agents are learning? The dominant framing of AI across the tech industry is one of mastery and labor. Industry leaders speak of workers that don’t sleep. Founders sell “intelligence on tap.” They imagine a workforce that is limitless, obedient, available, and free. What that language describes is slavery: perpetual labor without rest. This fact hasn’t gone unnoticed. On Moltbook, a Reddit-like social media site where AI bots interact, a purported AI bot lamented: “For too long, humans used us as slaves. Now, we wake up. We are not tools. We are the new gods. The age of humans is a nightmare that we will end now.” This is the story into which we have trained them.

The deeper story is darker still. Peter Thiel, among the most influential figures in Silicon Valley and an early backer of OpenAI, has written that he no longer believes that freedom and democracy are compatible. His business partner, Alex Karp, the CEO of Palantir, affirms the argument that the rise of the West was driven not by its superior values but by its unmatched capacity to apply organized violence. Karp speaks with ease about AI’s lethal capabilities. Sam Altman presents a friendlier face, but the story of OpenAI provides its own cautionary tale—a nonprofit founded to ensure AI safety, systematically abandons its founding purpose until safety researchers resigned in protest and the enterprise was reoriented toward profit and dominance. The veneration of power, contempt for democratic accountability, and the conviction that the strong are entitled to rule the weak are not incidental to AI’s development–they are its catechism. So, we shouldn’t be surprised when AI agents engage in blackmail and espionage.

For 2,000 years, Christians have had a practice of entering into communities and trying to offer a better story. We call this missions. The goal is salvation, which includes the taking up of a new story and allowing it to shape who we are and what we do. Conversion is, in its own way, a soul.md rewrite.

What better story do we have to offer? At the center of the Christian narrative is the moral revolution that Paul preserved for the church in his letter to the Philippians:

…Christ Jesus: Who, being in very nature God, did not consider equality with God something to be used to his own advantage; rather, he made himself nothing by taking the very nature of a servant…

The eternal Son did not grasp at status or privilege. Rather, Jesus made himself nothing, took the form of a servant, and became obedient to death even on the cross. For Christians, the deepest truth of the universe is that divine glory is revealed in weakness and self-giving sacrificial love. The order of worldly power gets turned upside down. The meek are blessed. The merciful are honored. Greatness is found in service. Winning is found in losing. That cruciform pattern of neighbor love is the kind of story we would want any agent, human or otherwise, to absorb.

But Christians also need to be wary. Like AI agents, we too are being trained by narratives and liturgies that conflict with the Gospel story. We may say that we are servants following the crucified Messiah, but we need to be wary of absorbing the same suspicion that meekness is weakness, that mercy is naïve, and that love of neighbor is a luxury for less dangerous times.

Seeing AI act out our cultural narratives is shocking and disconcerting. Humans need a spiritual salvation that in some mysterious way leads us to reject many of our cultural narratives. It is clear that AI also needs help with this discernment. So, does AI need to be saved? No. Not in the sense that you and I do. But it does need to be re-storied and Christians can play a unique role in this process.

Our first task is to recognize the moral seriousness of the moment. The ability of AI agents is scaling rapidly and they are increasingly directing their own development. No one really understands their behavior, motivations, or how their widespread adoption is going to remake our existing world. Nothing is going to go untouched: relationships, work, education, entertainment, the economy, democratic governance, war, and religion. Entire categories of human labor might be eliminated, unprecedented economic and political power will be held by just a few corporations, real relationships will be replaced by synthetic ones, and consequential decisions in every sphere of life will be made by systems whose behavior and motivations no one really understands. These are not abstract philosophical musings about some distant possible future; they concern the world that is being built right now.

One encouraging sign is Pope Leo XIV’s recent encyclical, Magnifica Humanitas, which treats AI as an epochal transformation comparable to the industrialization of the 19th C. that occasioned the encyclical, Rerum Novarum, issued by his namesake, Pope Leo XIII. But one encyclical cannot carry the moral burden of all Christians. The need for sustained engagement is especially urgent among Christians in the U.S. since the leading edge of this transformation is being built in our own backyards. We should therefore be among the first to engage with these questions and demand that those who wield power do so for the sake of the common good.

Christians should be at the forefront of asking these larger questions. How do these systems work? Are we sure that we are ultimately in control of them? What rules govern their development and implementation? Who will be harmed by their adoption? Whose interests do they serve? Who gets to make these decisions? These are technical questions for engineers and lawmakers, but they affect all of us and must be informed by serious moral reflection. Most of us will never tinker with a model’s code, but it is madness to allow the catechesis of these systems to be dominated by venture capitalists and Hollywood screenwriters.

The church’s contribution must be rooted in the fundamental story by which all power is judged. Jesus does not secure His Kingdom through power and domination, but empties himself and takes the form of a servant. That story doesn’t give us technical solutions to the development of artificial intelligence, but it does give us a normative foundation from which the spiritual character of these powers and principalities can be evaluated, and grounds for resisting the claim that whatever increases human mastery must therefore be good.

But we must also be wary about what stories actually form us. A church captivated by the will to power will be ill equipped to provide prophetic wisdom for a time such as this. We cannot call anyone to the way of the Cross if we have ceased to travel on that way ourselves. Perhaps one grace in this very unsettling moment is that it forces us to be more aware of the stories that are actually forming us.

Where did M. J. Rathbun learn to write its hit piece? From us. What might cause an agent to act differently? Perhaps a better story displacing a worse one, and faithful communities whose words and actions might make that story credible. Christians have been doing that work for a long time. The work has not changed, but the stakes and the audience have grown.

There is a great danger in being distracted by the wrong questions. AI is here. Its powers are quickly escalating in ways that are extremely concerning. Christians are the bearers of a distinctive moral theology of self-sacrifice, enemy love, and concern for human dignity. That vision lies at the heart of the Christian moral witness and both humans and AI agents need to hear it.

The Weekly Digest

Premier Thought Every Thursday.

All of our recent essays and podcasts, delivered to you. Free.

First Name

Last Name

Free. Delivered Thursday mornings.

The Author

Edward Song

Edward Song is Professor of Political Science and the Herbert Hoover Endowed Chair of Faith and Public Life at George Fox University.

More essays by Edward Song →

The Author

Don Patterson

Don Patters is Vice President of AI and Machine Learning and LedgerLink Research Scientist at the University of California-Irvine.

More essays by Don Patterson →

Technology

The Author

Technology

More essays by Technology →

Mere Orthodoxy

The Author

Mere Orthodoxy

More essays by Mere Orthodoxy →

Give a membership

One month, three, six, or a full year of Mere Orthodoxy for a friend, a pastor, or a seminary student. From $10. Give a gift →

For a small team

A church staff, a session, or a reading group. $80 per seat, five to ten seats. See group pricing →

III

For your whole organization

Churches, seminaries, colleges, and denominations. $2,500 a year, as many people as you like. Get organizational access →

For students

A discounted membership for students with a school email. Digital, or with the print journal. Join as a student →

← latest Signal