Briefly in Defense of The Gospel Coalition's AI Policy
Aim small, miss small.
The Law of Diminishing Restriction
Stipulate the following:
Guidelines about AI use are an absolute good, both at the personal and the institutional level.
Tech guidelines are in general unpopular and difficult to maintain over time.
Thus, most guidelines about AI use will, directionally at least, “broaden” rather than “narrow” over time. In other words, if an institution’s AI policy in 2026 is “Try not to do X,” the odds are good that by 2029 the policy will be “Doing X is OK but try not to do Y.”
Cultural examples of #3 with regard to smartphones are everywhere. Fifteen years ago, most churches and classrooms did not bother addressing whether phones were welcome, but you did get scolded if your phone went off. Today, everyone accepts that you will be on your device. Can you at least mute it? Yes, you can text right outside the concentration camps; can you at least not snap a selfie?
Technology is not linear (don’t let anyone tell you it is). But the law of diminishing restriction is pretty reliable. Most technologies get their strictest regulation and stigmas in their very early years, for the simple reason that not enough people are demanding access yet. Once the controls start falling, they fall quickly and in batches.
This is why I think The Gospel Coalition’s recently published AI policy is, on the whole, pretty good.
To be fair, I think the category of “research” requires more nuance than their published essay gives. That aspect of the policy seems like it doesn’t quite take seriously enough the changes that Google and other online engines are making to fully integrate AI into their results. It seems implausible that a regular person, without expensive access to paid databases or university libraries, would be able to utilize a search engine without experiencing some kind of AI-generated help.
But let’s put that aside. The reason the policy looks good to me is that it draws lines around AI and creative work that will, by virtue of their harshness, resist the law of diminishing restriction. This is an AI policy intended to alienate those who want to reap the benefits of not doing the work. The policy isn’t meant to make it easier for creators to get read at TGC. It’s meant to make it easier for readers to take seriously what they’re reading.
Aim Small, Miss Small
Here’s a quote from Brett McCracken:
We’re not looking for perfect writing. We’re not looking for meticulously organized information full of suspiciously pithy one-liners or vague examples that don’t feel like they originated in a human’s real life.
We’re looking for writing that comes from a human heart, born out of real passion, nagging questions, persistent burdens; writing characterized by human affection more than perfectly calibrated linguistic inflection. We delight in the hallmarks of truly human writing: a bit messy, unexpected in the way real life always is, somewhat scattered and less efficient than machine-rendered words—yet pulsating with life.
To be clear, this doesn’t foreclose the possibility that genuinely human writing can stay human even while getting help from nonhuman tools. It rules out the possibility that people who see zero intrinsic value to human writing can produce writing of value.
The bar is worth setting here for several reasons. First, the AI revolution is a question of human identity before it’s a question of human ethics. By that I mean that the most important question anyone needs to ask when they’re thinking about AI is not, “Am I allowed to do this,” but, “What am I, and what is this?” The law of diminishing restriction starts by getting people to forget to ask. There is only one path toward coexisting with AI technology in a non-destructive way, and it goes through this question.
Already there have been folks online bewailing TGC’s policy for being too strict. I’ll be honest: I don’t know how to stake out common ground with people who have no interest in asking the existential question with regard to tech. I’m perfectly comfortable living and working with people who come to different conclusions. But people who believe there is no meaningful reflection to be done are taking themselves and others into a dangerous corner. People who don’t care about a meaningful anthropology, who don’t do any work to answer the question, “Why replace humans this way, but not that way,” are just volunteering their own abolition.
Second, the operative principle here is “Aim small, miss small.” If a marksman gives himself a wide target area, that might feel like increasing his chances of success, but it’s also increasing his chances of complete failure. Aiming for any of the target means if you miss, you’ve hit none of it. On the other hand, if you aim for the bullseye, you may miss, but it’s far likelier you’ll end up hitting something. An overly strict AI policy will miss, but it will miss in a better way than the alternative. It will miss in a way that still gives people the spiritual categories to understand who and what they really are. It will miss in a way that preserves human agency and relationships. And again: this is all contrary to the force of cultural gravity. Our relationship with technology will, all variables being equal, slacken, not sharpen, over time.
A Not-Normal Revolution
If restrictions seem strict, consider how extreme our experience of the technology really is. Is it not insane that we can summon entire book-length pieces of content for no other reason than we asked for it? Is it not insane that people can develop intimate and romantic relationships with voices and icons, whose output would fool the greatest minds in world history? Is it not insane that we need not go anywhere, touch anything, or learn from anyone, in order to “create” whatever we can imagine? It’s insane. Doesn’t such a “not-normal” normal demand at least some meaningful reflection about it? How severe are such restrictions really, compared to the power of the tool?
Here’s my point: If you’re not actively thinking about what this tech cannot be or do, you’re not being “realistic.” You’re being transhuman. You’re taking less than five years of innovation as a new baseline of human existence. You’re moving the goalposts of your own personhood. But you don’t have to. You don’t have to have all the answers or the perfect plan. You just have to really do what AI only pretends to do: Think.