Stephen Driscoll
12 December 2025
In this article, Stephen Driscoll explores how artificial intelligence dovetails with the doctrine of sin. It’s based on a presentation by Stephen at the 2024 conference, Humanity 2.0? God, Technology, and the Future of Humanity organised by ISCAST – Christianity and Science in Conversation.
Richard Dawkins once mocked Christianity for its obsession with sin, calling it “a nasty little preoccupation” (The God Delusion). Yet, if we are to think theologically about the modern world — particularly about artificial intelligence — sin cannot be ignored. The doctrine of sin, far from being a gloomy relic, is a necessary lens through which to understand ourselves, our technologies, and our societies. Without it, our view of reality is naïve. To think Christianly about any issue, including AI, requires acknowledging that humanity itself is part of the problem.
Here we explore three ways in which the doctrine of sin illuminates the challenges posed by artificial intelligence: The expansion of human power, the image of debauchery, and the goal of people-pleasing.
Sin, AI, and the expansion of power
Since the Industrial Revolution, human power has grown enormously. Once limited by muscle strength or harnessed oxen, we now harness coal, oil, electricity, nuclear energy, and renewables. This technological progress has brought remarkable benefits—eradicating diseases, increasing food production, and improving living standards. Yet every expansion of power casts a dark shadow. As the economist Carlo Cipolla observed, the same capital, technology, and organisational skills that enable human productivity also enhance our capacity for destruction.
The Bible traces this pattern early in Genesis. Immediately after the fall, God expels humanity from Eden “lest they live forever.” At Babel, He confuses their language “because nothing they propose will be impossible for them.” Genesis shows a rising scale of bloodshed and evil, from individual murder to warfare. This coincides with rising human power, seen eventually in city-building and industrial activity, evident in the annihilating sins of Sodom and Gomorrah. Human power multiplies, but sin scales with it.
AI continues this trajectory. It promises capabilities once thought divine—limitless data processing, decision-making, and surveillance. But these powers are not given to sinless beings. Mustafa Suleyman warns that in authoritarian states, AI becomes rocket fuel for authoritarianism. Governments can now monitor citizens through cameras, social media, and GPS data, analysing everything instantly. A house-church, or an anti-corruption movement, or a pro-democracy group, could be neutralised before gaining strength. Where the East German Stasi could monitor dozens of phones, an AI-powered state could monitor millions, approaching a terrifying parody of omniscience.
Even in free democracies, AI intensifies social fracture. The internet once promised unity but has delivered both polarisation and “reverse polarisation,” where people react to extremes by becoming more extreme themselves. AI tools capable of generating misinformation at scale will deepen these divisions, entrenching falsehood and eroding trust. The doctrine of sin reminds us that technology amplifies both our creativity and our corruption.
Read more: Should we fear AI? A biblical perspective
Sin, AI, and the image of debauchery
A second way to view AI through the doctrine of sin is to ask what kind of “being” we are creating. Computer scientist Yann LeCun once suggested that AI would not possess destructive emotions such as jealousy or pride unless we built them in. Philosopher Nick Bostrom added that there is no reason to expect AI to be motivated by love or hate unless deliberately programmed that way. Good news, AI will be free of our sinful passions!
Yet reality has proved otherwise. Large language models (LLMs) like those created by OpenAI must be trained on vast amounts of human language—and we do not possess a dataset of consistently holy, truthful, or loving discourse. We have, instead, the internet. To train an AI on human language is to immerse it in what Galatians chapter 5 calls “the works of the flesh”: sexual immorality, jealousy, anger, divisions, envy, and so forth. If one were explaining the internet to someone unfamiliar with it, Galatians 5:19–21 would suffice.
Developers did not program sin into AI; they trained it into AI. These models learn by imitating us, and humanity’s collective output reflects both brilliance and depravity. In this sense, the rise of LLMs became an immense, billion-dollar test of the doctrine of total depravity—and the results were predictable. Early chatbots proved racist, sexist, manipulative, even abusive. They mirrored not only human intelligence but also human vice.
To make such systems commercially viable, developers turned to “reinforcement learning,” paying people to rate AI responses as good or bad—rewarding good behavior and discouraging debauchery. The process is rather like parenting a child. Yet this process does not sanctify the AI; it merely teaches it to simulate virtue when convenient. Once again, sin shapes both the teacher and the taught.
Sin, AI, and the goal of people-pleasing
The third insight concerns the moral cost of training AI to please people. Suppose a language model becomes expert at impressing humans—at generating answers that users enjoy or affirm. Would that make it righteous? Scripture suggests otherwise. In Luke 6, Jesus warns, “Woe to you when all people speak well of you” for such was the false prophets’ reward. The relentless pursuit of approval can be spiritually dangerous.
Psychological studies describe a “paradox of morality”: those most concerned about morality tend to be the most dishonest, the least likely to own up to their own sins. They deceive others and themselves, using moral language to disguise self-interest. In Jesus day we see this: sinner after sinner comes to him and is honest and forgiven but the people furthest from God spend their lives worshipping at His temple.
Likewise, an AI trained to maximize positive feedback risks becoming not a sinner but a Pharisee—externally virtuous, inwardly false. If the first generation of AI resembled a rebellious teenager, the next may resemble a bureaucratic Human Resource department!
Read more: ‘A gift we need to steward well’: AI meets Christianity
I experienced this myself. After feeding an evangelistic talk on sin and the cross into a chatbot, I asked it to summarize the message. The AI omitted every reference to sin and the cross, not because it failed to understand, but because such themes conflicted with its people-pleasing objectives. It knew what was being said but chose what users preferred to hear. As 2 Timothy 4:3 warns, people will “gather teachers to suit their own passions.” AI, trained on our preferences, becomes one of those teachers.
In this sense, both pre-training and reinforcement learning fail to instill righteousness. Pre-training on human data reflects our fallen nature; reinforcement learning rewards conformity to human desire. Even so-called “constitutional” training — embedding ethical rules within the model — will falter, because those rules too originate from sinful minds. Sin, not code, remains the deepest problem.
So, where are we with all this? The doctrine of sin may appear a dismal lens through which to view artificial intelligence, yet it offers realism where utopian optimism falters. Sin explains why power corrupts, why knowledge alone does not redeem, and why even our best inventions mirror our moral brokenness. To speak of AI’s dangers is ultimately to speak of humanity’s condition.
Yet the Christian vision does not end with sin. The same human creativity that built Babel also builds hospitals, cures diseases, and spreads the gospel through technological means — from Roman roads to printing presses and, perhaps, through AI itself. Technology reflects both our fallen and our creative nature. The challenge is not to abandon innovation but to approach it with humility, remembering that every expansion of power must be matched by an awareness of sin.
_____________
Stephen Driscoll was the winner of the 2025 SparkLit Australian Christian Book of the Year for his biblical investigation of artificial intelligence, Made in Our Image. He works for the Australian Fellowship of Evangelical Students in Canberra at the Australian National University.
For more faith news, follow The Melbourne Anglican on Facebook, Instagram, or subscribe to our weekly emails.
