The consensus narrative is comfortable and familiar: artificial intelligence is a tool. Like the printing press or the internet, it will disrupt some jobs, create others, and eventually settle into our lives as another utility we can't imagine living without.
This framing lets us off the hook. It suggests that AI is something external we can manage, regulate, and control. We can have hearings about it. We can write rules. We can retain our decision-making authority while letting the machines handle the grunt work.
The actual problem is messier and more corrosive than that.
What we're witnessing isn't just the adoption of a new tool. It's the outsourcing of judgment itself, and the infrastructure that supports human judgment is fracturing faster than our institutions can acknowledge.
Consider what happened in recent weeks: a criminal referral that raised questions about whether generative AI helped draft official government documents. A social media ecosystem increasingly populated by AI-generated content that humans struggle to distinguish from reality. Institutions from newsrooms to congressional offices quietly deploying language models to draft communications, analyze documents, and shape arguments. Judges and prosecutors beginning to rely on algorithmic recommendations without fully understanding how those systems work.
The comfortable story says: "This is fine. We'll figure out the guardrails." But that misses what's actually breaking.
The first thing breaking is institutional memory. When a prosecutor uses an AI tool to analyze case files, when a legislator relies on an algorithm to draft policy language, when journalists use automated systems to surface stories, these institutions stop developing the human expertise that once made them reliable. You can't get that expertise back quickly. It atrophies. The next generation doesn't learn how to do it the hard way because the hard way is now considered inefficient.
The second thing breaking is accountability. If a judge considers an algorithmic risk assessment in sentencing, where does responsibility lie when that assessment is biased or flawed? If a government agency uses AI to make decisions affecting citizens, how do you audit what happened? The tool becomes a convenient explanation that shields the humans behind it. "The algorithm recommended it" is not accountability. It's diffusion of accountability.
The third thing breaking is the very concept of professional credibility. We once trusted lawyers, doctors, journalists, and civil servants partly because they had internalized standards of their profession. They made human judgments informed by training, experience, and ethical obligations. When those professions become primarily about managing and interpreting AI outputs rather than making independent judgments, the profession itself changes in ways we don't fully understand yet.
The comfortable consensus says regulation will solve this. But regulation assumes we understand what we're regulating. We're making rules for systems that are, by design, partially opaque to the people using them. We're writing guardrails while the car is already in motion and accelerating.
None of this means AI is inherently evil or that the technology shouldn't exist. But it means the "tool" metaphor is doing a lot of work to make us comfortable with something more fundamental: the substitution of algorithmic judgment for human judgment, not as an exception but as a default.
The real question isn't whether AI will disrupt employment or whether we need better guardrails. Those are important questions, but they're comfortable ones that let institutions feel like they're responding.
The real question is: what does a society look like when the institutions that were supposed to exercise independent judgment have systematized that judgment away? What happens to institutional credibility when nobody can fully explain how decisions were made? What breaks next when we've outsourced thinking to systems we don't fully understand?
Those are the questions our consensus isn't asking yet.