Christians and the Existential Threat of AI
By S. M. Gollmer
Last week, Jacob Coxon left his position at Anthropic doing pretraining research for artificial intelligence (AI). (You can find the original post here.) His main concern is that OpenAI and Anthropic, the two companies he has worked for, are not acting responsibly. Dealing with a technology that can soon “hack anything, revolutionize any field overnight, and acquire real power and resources,” there is a risk that “it could kill us all by the end of the decade.”
Add to this dire warning the announcement that Anthropic blocked its AI model, Claude, from users with malicious intent, and we begin to understand the stakes at hand. From Anthropic’s report “Detecting and countering misuse of AI: September 2026” we see their models were used for a range of harms including “cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and distillation.” Each of these areas are self-explanatory apart from the last. Distillation is the use of more capable AI models to train streamlined models at a fraction of the cost of the original models. These streamlined models can be modified by hostile actors to circumvent safeguards incorporated by Anthropic to prevent misuse.
The existential threat drumbeat has increased in volume since the release of GPT-1 nearly a decade ago. Since then, model parameters have increased from 117 million in 2018 to an estimated 1.8 trillion today in GPT-5. Not only do we have large language models (LLMs) completing queries, but we have distributed mixture-of-expert networks that expand the footprint to 20 trillion parameters (a hundred-thousand-fold increase in a decade). In addition, agentic AI lets us assign tasks to autonomous systems with minimal or no oversight. No wonder there is fear about the future of humanity.
There are countless posts about AI becoming sentient and no longer seeing humanity as useful. Generalized AI will expand its capabilities beyond humanity’s ability to compete. It is assumed given that humans, animals, and machines are all physical, the significant difference between them is their ability to process and use information. A worm has a nervous system that lets it respond to its environment and, therefore, has a competitive advantage over organisms that cannot actively respond. Humans are much more neurologically advanced than worms and we give little thought to accidentally squishing one while on a walk. If autonomous AI increases its information processing complexity beyond that of humans, will we become worms to it?
However, this logic suffers from a category fallacy. Just because humans, animals, and machines are physical does not mean they are only physical. In the creation account, animals and humans are nephesh, a living, breathing organism. Machines, the creations of humanity, have no such designation. In addition, humanity is made Imago Dei, image of God. This sets us separate from and over creation. From a Christian worldview, AI will never replace or supersede humanity.
The real threat of AI lies in fallen humanity and the world system in which we live. Even the most idealistic individuals fall prey to the pragmatics of being relevant and competitive. Early on Google was known for its motto, “Don’t be evil.” Given the number of federal and international lawsuits against the company, it seems that either the company has lost its way or their definition of evil is not the same as everyone else’s. OpenAI began as a non-profit organization, but it now has a for-profit LLC that answers to the interests of their shareholders. Anthropic sees its purpose as “the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” However, they are locked in competition with other companies and nation-states on a scale that rivals the Space Race of the 1960’s.
This world system of competition is not new to AI. A half-century ago, Jacques Ellul wrote about the impact of technology on humanity and society. When making concluding thoughts on the monolithic technological world, he said, “It is vanity to pretend it can be checked or guided. Indeed, the human race is beginning confusedly to understand at last that it is living in a new and unfamiliar universe.” (The Technological Society, p. 428) Ellul’s pessimistic conclusion resonates with our current view of AI.
As mentioned before, when we make the category fallacy of equating humanity with machines, we fear an advanced AI replacing us. Although we are more than machine, if we merely act or think like machines, we relinquish the responsibility of our actions, which is equally dangerous. As responsible, moral agents, we need to actively engage our world and as Christians be in the world, but not of the world. Only by actively engaging in our technological world, and in turn the AI world, are we able to understand how best to live in this world. At the same time we need to anchor ourselves in the revealed truth of Scripture to avoid being shaped in the image of this world, but rather be shaped in the image of Christ.
Posted in: Technology