Grappling with the AI Threshold

A society that concludes that humans are valuable in proportion to their economic output will use AI to eliminate the humans it no longer needs.

David Cassady

22 Apr 2026 • 8 min read

I use artificial intelligence nearly every day. It drafts with me, argues back, helps me work through difficult memos, and occasionally produces something I could not have produced myself in any reasonable amount of time (slides, diagrams, summary charts, etc.). I am not apologetic about this. Like anyone learning to work with a new instrument, I am figuring out what AI does well, what it does poorly, and what it should not be asked to do at all.

And yet I am also a Professor of Media Studies, which means I am paid — quite literally — to not take the tools of our moment at face value. Every powerful medium in human history has been sold to us with a story about freedom and progress, even when the effects on actual human beings proved more complicated. Television was going to educate us all. The internet was going to flatten hierarchies. Social media was going to connect everyone. Each of those stories had some truth and a great deal of marketing. AI is now being sold to us with a similar story, and we owe it to ourselves — and to our neighbors — to listen with discernment.

So let me lay my cards on the table. AI is part of reality now. Like every technology before it, it can be used for good and for ill, and it almost certainly will be used for both. That is not pessimism; it is the plain record of what happens when powerful tools become widely available. According to Pew Research, about one in five U.S. workers already use AI as part of their jobs, up from one in six a year earlier.^[1] The question is no longer whether AI arrives. It has arrived. The question is what kind of arrival we are willing to accept.

The questions AI is raising are not, finally, about AI

Underneath every debate about artificial intelligence — the boosters, the doomers, the careful academics, the pastors wondering what to preach — lie a handful of very old questions.

What is a human being worth? Is a person valuable because of what they can produce, or because of something that is innate in their being... in their creation? What is the purpose of work? What is the role of society in encouraging human thriving?

Those are not technology questions. They are moral and theological questions, and AI simply puts them in front of us again with new urgency.

In January 2025, the Vatican released a document called Antiqua et Nova, a thoughtful reflection on the relationship between artificial intelligence and human intelligence. It is worth reading in full, but one sentence has stayed with me: "A person's worth does not depend on skills, cognitive achievements, or individual success, but on the person's inherent dignity."^[2] That is not a Catholic claim alone. It is an ecumenical claim, a deeply Christian claim, and it sits in direct tension with a cultural narrative that increasingly treats productivity as the measure of personhood.

The theologian Noreen Herzfeld, who teaches both theology and computer science, has argued for years that the imago Dei — the claim that human beings are made in the image of God — is primarily about relationship, not cognition.^[3] We are not image-bearers because we can reason or calculate. We are image-bearers because we are made for one another and by God. If that is right, then no chatbot, no agent, no humanoid robot on an assembly line can ever be a substitute for a human being. It can be a tool, and sometimes a remarkably good one, but it cannot be a neighbor.

This is exactly where AI presses us. A society that quietly concludes that humans are valuable in proportion to their economic output will use AI to eliminate the humans it no longer needs. A society that insists on the a priori, irreducible worth of every person will use AI differently. Those are not the same society, and they will not treat the vulnerable in the same way.

Things are developing quickly — and that matters

I want to be careful here, because the most common failure mode in writing about AI is to sound either breathless or dismissive. The truth, as nearly always, is in between.

The pace of development in 2026 has been, by any reasonable standard, astonishing. We have moved beyond "chatting with an AI" to something the industry now calls agentic AI — software that can take actions on your behalf. An agent can book your travel, fill out a form, work through a multi-step research task, or draft and send an email. Anthropic and OpenAI have both released consumer-facing agents in the last year. Whatever you think of this development, it is no longer theoretical.

Robotics has moved in parallel. The same companies racing to build smarter language models are also building humanoid robots, and the two projects feed each other: a robot body needs a mind, and an AI mind gets much more interesting once it can pick things up, open doors, and respond to a physical environment. Tesla, Figure, Unitree, and a number of Chinese firms are all now shipping or mass-producing humanoid units; 2026 has been widely described as the commercial inflection year.^[4]

Here is the sobering part. This pace is not an accident. It is the product of a race.

The most aggressive AI development in the world is happening between American and Chinese firms. As of this spring, the performance gap between the best American and the best Chinese frontier models has reportedly collapsed to about 2.7 percentage points — even though the United States spent roughly twenty-three times more on private AI investment last year.^[5] That is a stunning fact, and it should make us all uncomfortable, though perhaps not for the reasons we first assume. Races of this kind tend to produce shortcuts.

There is already evidence of exactly that. Reporting from inside several major labs indicates that internal "safety thresholds" — the capability levels at which enhanced safeguards are supposed to kick in — were revised upward at least four times between January 2024 and December 2025. In other words, when models in development kept exceeding a company's own stated safety limits, the limits got raised rather than the deployments halted. Executives cited competitive pressure from rival labs.^[6] That is not ill will. That is what races do to good people. They push thoughtful engineers into decisions their better selves would not make.

And so far, governments have largely chosen to enable the race rather than to regulate it. In December 2025, the White House issued an executive order instructing the Department of Justice to challenge state-level AI laws in court and to identify state regulations the administration considered "onerous."^[7] Federal broadband funding was tied, in part, to states falling in line with the administration's AI preferences. Whatever one thinks about the appropriate balance between federal and state authority, that is a posture of clearing the track — not of slowing the runners down.

But we get a say

This is the part of the conversation I most want to protect, because it is the part most easily lost.

The fact that AI companies now spend enormous sums on lobbying and public messaging should tell you something. In the first quarter of 2026 alone, Anthropic more than tripled its lobbying budget from the same period a year earlier; OpenAI nearly doubled its own; the broader tech sector poured more than a hundred million dollars into federal influence in 2025.^[8] Companies do not spend that kind of money unless they believe the public's opinion — and the government's response to that opinion — can actually constrain them.

In other words: what we think matters. What we say to our elected officials matters. What we ask of the firms we work for, buy from, and teach about matters. The sheer intensity of the messaging is itself an admission.

The same is true of the quiet campaign to shift AI's costs onto the rest of us. Data centers now account for roughly half of all new U.S. electricity demand,^[9] and residential electricity prices have risen nearly fifty percent in just a few years. In response, a group of major AI firms recently signed what they called a "Ratepayer Protection Pledge," in which they promise to cover the new generation costs their data centers require. Critics have fairly described that pledge as "a pinky swear with no enforcement mechanisms."^[10] Meanwhile, communities in Virginia, Arizona, and elsewhere are watching their power bills climb and their water tables drop so that a company in another state can train another model.

Here, too, voices and votes matter. Public utility commissions, local zoning boards, state legislatures, and Congress all have real levers. They will not use them on their own. They will use them if we ask.

And Americans are, in fact, beginning to ask. Pew reports that half of U.S. adults now say AI makes them more concerned than excited, and only about one in ten feel the reverse.^[11] That concern crosses party lines in a way almost nothing else does right now. It is not a sign of panic. It is a sign of a public waking up.

A threshold season

In the Christian tradition, a threshold is not simply a boundary. It is a place where decisions are made. A door is opened or not. A guest is welcomed or not. A direction is chosen.

That is what I want to say to my colleagues in ministry, my students, and my fellow citizens: we are in a threshold season with respect to AI, and our voices and our choices still count. The future of this technology — whether it serves human flourishing or erodes it, whether it dignifies workers or discards them, whether it cares for the creation that sustains us or degrades it — will not be decided by engineers alone, and certainly not by executives alone. It will be decided, in part, by what the rest of us demand.

I use AI nearly every day. I will go on using it. But I refuse to pretend the wider questions are someone else's problem. The church, the academy, the voting booth, and the family dinner table are all places where AI is going to be reckoned with in the years ahead. Our faith, our ethics, and our attention have a real role to play.

The questions AI is raising are not, finally, about AI

Things are developing quickly — and that matters

But we get a say

A threshold season

Sign up for more like this.