AI May Kill Us All. Meanwhile It’s Giving the Usual Suspects a Leg Up.

Photo by Guillaume QL on Unsplash

The discourse on AI is dominated by big questions. Whether superintelligence will eventually slip the leash and treat humanity as an obstacle. Whether the data centers will accelerate climate collapse. Whether the models will displace whole categories of work in a decade or in a generation. These questions are serious and they deserve serious people working on them. The cataclysmic scenarios are not silly. They are hypothetical and distant, possibilities arrayed across a horizon that may or may not arrive.

What is neither hypothetical nor distant is what AI is doing right now, on this Tuesday, in this country, to people whose lives are already shaped by every previous wave of technological sorting. The cataclysm gets the headlines. The local, immediate, already-happening exclusions get a footnote. AI is doing what most powerful technologies have done before it. It is accelerating the people who already had the wind at their backs. And it is leaving behind, again, the people who have always been left behind.

This is the conversation that isn’t happening. It is happening in pieces, in academic papers about algorithmic bias and threads about generative models that can’t draw Black hands correctly, but the larger pattern has not yet been named. So let’s name it.

The split is not who can use AI. It’s who can trust it.

A generation of professionals is right now using AI fluently and constantly. Knowledge workers under fifty, mostly, with secure jobs and reasonable confidence in their own competence. They draft with it. They code with it. They think with it. They are shipping more, charging more, and pulling further ahead. Consulting firms whose throughput would have required three additional employees five years ago now run lean and fast on a single operator with a model.

A second, larger group has been told, often by the same media ecosystems and the same employers, that AI is dangerous, dishonest, plagiarizing, environmentally ruinous, and probably going to take their job. They have been told this is the moral position. They have absorbed it. So they are not using the tools, or they are using them apologetically, in secret, with a layer of guilt that prevents them from getting good at it.

The first group is racing ahead. The second is being told that hesitation is a virtue. The first skews wealthier, whiter, more credentialed, more male, more coastal. The second is going to look up in three years and discover that the gap has become structural.

Fear is being distributed unequally, and so is permission.

Fear of AI is real and some of it well-justified. Surveillance harms fall harder on Black and brown communities. Generative models reproduce stereotypes. Voice cloning enables new kinds of fraud aimed at elders. Content moderation systems silence queer creators while letting harassment through. None of this is paranoia. It is the historical record updated to the present tense.

But notice what the fear-narrative does in practice:

It tells the people most likely to be harmed by AI that they should also be the people least likely to learn it, use it, shape it, or profit from it. The result is a double tax. The downside risk lands on you. The upside passes you by.

Meanwhile the people who are least likely to be surveilled, deepfaked, or stereotyped by these systems are the ones being handed enterprise licenses, training budgets, and the social permission to experiment in public. They get to be early adopters. Everyone else gets to be cautionary tales.

What the gap costs.

The people leaving the most productivity on the table are often the people doing the most important work, because they have decided that using AI would be a moral failure. A nonprofit director writing grant reports by hand at midnight. An adjunct professor cobbling together three classes worth of lesson plans without help. A first-generation lawyer billing hours she could halve. A community organizer transcribing interviews that a model could caption in seconds. A public defender researching case law on a caseload built for three of her.

The cost is not just personal. It is generational. Every hour these people spend on tasks that AI could absorb is an hour they are not spending on the work only they can do, the relational work, the political work, the creative work that requires their specific judgment and lived experience. The technology that could have given them leverage is instead giving leverage to the people already leveraged.

And income compounds. Career velocity compounds. The 2026 entry-level worker who learns to direct AI well is going to be running circles, in five years, around the equally-talented peer who was warned off the tools by a well-meaning professor in 2024. That gap will not close on its own.

The honest conversation.

Two things have to be held at once. AI is genuinely dangerous, in ways both spectacular and mundane, and the marginalized will absorb a disproportionate share of those harms. And, AI is genuinely useful, in ways that compound across a career, and the marginalized are currently being talked out of accessing those benefits by people who already have them.

Both of these are true. Pretending only the first one is true is a luxury position. It is a position available to people whose jobs, incomes, and trajectories are not the ones being optimized away.

What this moment calls for is a refusal of the false choice. The same communities that are organizing against algorithmic harm should be the communities organizing access to AI literacy, to the capacity to build with these tools rather than only defend against them. Public libraries running AI workshops. Union contracts that include AI training as a worker right rather than a management threat. Fellowships that put models in the hands of organizers, journalists, and artists from outside standard pipelines.

The cataclysm conversation can keep happening. It should keep happening. But it cannot be the whole conversation, because while it absorbs the oxygen, something is unfolding at ground level that is sorting people. Quietly. Predictably. Along exactly the lines you would expect.

The Terminator threat is worth watching, but from a telescope. The threat that needs regular glasses, the one already in the room, is the same pattern visible in every wave of technological change since the open web. A small group capturing the upside. A much larger group absorbing the downside. And a discourse so fixated on the dramatic and the distant that it cannot see the ordinary unfolding of advantage right in front of it.