Our Data. Our Resources. Our Terms and Conditions.

The neuroscience of learned control, the AI safety crisis no one should ignore, and why the fight over data centers is really about who holds the power.

In 1967, a young psychologist named Martin Seligman ran an experiment on dogs that became one of the most famous in the history of psychology. Dogs were placed in a chamber and given electric shocks they could not escape. Later, those same dogs were moved to a new chamber where escape was easy. All they had to do was jump over a low barrier. But the dogs that had previously experienced inescapable shocks didn’t jump. They laid down. They accepted the pain. They sadly whimpered and gave up.

Seligman called this learned helplessness, and the concept shaped how we think about depression, poverty, and human motivation. The idea was that organisms who learn they have no control over bad outcomes stop trying, even when the situation becomes changeable.

Most people who know anything about psychology know this finding. What most people don’t know is that fifty years later, Seligman himself said he got it backwards.

The Update That Changes Everything

In 2016, Seligman and his longtime collaborator Steven Maier published a paper in Psychological Review called “Learned Helplessness at Fifty: Insights from Neuroscience.” Five decades of brain research had converged on a finding that inverted the original theory. As they wrote, “the original theory got it backward.”

As it turns out, passivity in response to prolonged aversive events is not learned. It is the default. It is the unlearned, baseline mammalian response, mediated by serotonin activity in a brain-stem structure called the dorsal raphe nucleus. When something painful happens and nothing you do stops it, your brain’s evolved default is to give in and just endure it. This is the factory setting. It is the reaction of a brain that has never had reason to believe that effort matters.

It is control that needs to be learned.

The prefrontal cortex, the most recently evolved part of our brain, is where the learning of control happens. When an organism experiences a situation where its actions actually make a difference, the prefrontal cortex learns to recognize that causal relationship. It then sends inhibitory signals down to the dorsal raphe nucleus, overriding the default passivity. The experience of control acts as an immunization against helplessness. Animals that first experienced controllable stress and then were exposed to uncontrollable stress didn’t give up. Their prefrontal cortex had already learned that effort can change things, and it applied that expectation even in a new and different context.

The experience of control in one situation protected against helplessness in a completely different situation. It generalized. It became a kind of cognitive vaccine.

I believe this is an extremely important finding, and I think it has enormous implications for politics, for humanity, and for the fight that is right now happening across America over data centers.

The Default Is Helplessness

Think about what this updated research means for how we understand human behavior. The default, biologically, is not agency. It is not initiative. It is not bootstrapping. The default is passivity in the face of prolonged pain. The ancient circuitry in our brains, the circuitry we share with every mammal, says: when things hurt and nothing you do seems to stop the hurting, just stop trying.

Agency, the feeling that our actions matter, is something that has to be learned. It requires experience. It requires actually having an effect on the world around you and recognizing that you did that. Without that experience, we accept the suffering.

Now look around at our “democracy.”

Look at a country where a landmark study by Gilens and Page found that “the preferences of the average American appear to have only a minuscule, near-zero, statistically non-significant impact upon public policy” once the preferences of economic elites and organized interest groups are controlled for. The average person’s opinion, year after year, has essentially no measurable effect on what the federal government does.

People feel this. They may not have read that study, but they know it in their bones. They feel the electric shocks of a system that doesn’t respond to them, and over time, many of them do what the dogs did. They just stop trying.

This is not laziness. This is not apathy. This is not ignorance. This is the default response to a loss of control, and it is operating at the level of the brain stem.

What We Do With the Pain

However, when human beings experience stress they feel powerless to address, they don’t just sit there quietly. The stress has to go somewhere. It gets processed through whatever channels are available, and most of those channels are pretty unhealthy.

Psychologists have a name for one of the most pervasive of these channels. It’s called System Justification Theory. Developed in 1994, SJT describes a motivated psychological tendency to defend, justify, and rationalize the existing social, economic, and political order, even when that order is actively harmful to you. The theory’s most counterintuitive finding is how under conditions of threat and stress, people at the bottom of a hierarchy often defend the hierarchy more vigorously than people at the top.

Why would anyone defend a system that hurts them? Because defending it reduces stress. This is referred to as the “palliative function” of system justification. Believing the system is fair, that people generally get what they deserve, that the way things are is the way things should be, provides what amounts to psychological relief. It reduces the painful cognitive dissonance of knowing you’re being harmed by something you feel powerless to change. If you can’t change it, the next best thing your brain can do is decide it doesn’t need changing, because actually it’s a good thing.

This is the Just World hypothesis at work. It’s the belief that the world is basically fair. That good things happen to good people and bad things happen to bad people. That poverty is earned and wealth is deserved. And it is, as I’ve written about before, a massive fallacy. But it persists because it’s psychologically useful. It turns the pain of injustice into the comfort of acceptance.

System justification isn’t the only way people cope with uncontrollable stress. Conspiracy theories are another. When people feel a loss of control over their environment, they are significantly more likely to endorse conspiracy theories, which offer the compensatory illusion of understanding hidden forces. A 2022 study in PNAS by Marten Scheffer and colleagues found that stress increases the rigidity of the beliefs underlying psychiatric disorders, prejudices, and conspiracy theories, and that this creates a self-reinforcing cycle. Rigid beliefs lead to worse decisions, which create more stress, which further strengthens beliefs. The authors suggested that “measures reducing social stress, including economic policies such as universal base income, may be the most effective ways to counteract this vicious cycle.”

There’s also displaced aggression. When people are stressed by forces above them in a hierarchy and feel unable to push back upward, they push downward. They bully those with even less power. They scapegoat immigrants. They lash out at their spouses and kids. They punch down.

And there's drug abuse. When pain becomes chronic, people self-medicate. Drugs provide immediate relief from stress that nothing in a person's environment seems capable of addressing. This is also why one of the most consistent findings from cash transfer research is that giving people money does not increase drug and alcohol use but actually tends to reduce it. A World Bank meta-analysis by Evans and Popova examined 19 studies across multiple continents and found that almost without exception, cash transfers had either no effect or a significant negative effect on spending on alcohol and tobacco.

And then there's religion. I say this as an atheist who respects what religion does for people. It works, at least in part, because it provides a sense of control and community. Both reduce stress. Believing that a higher power has a plan, that suffering has meaning, that you belong to a community of fellow believers, all of this lowers the cortisol load. Tellingly, when psychologists experimentally lower people's sense of personal control, people lean harder on both God and government. Religion and system justification fill the same void.

Every one of these coping mechanisms, from system justification to conspiracy theories to displaced aggression to organized religion to drug abuse, is a response to the same underlying problem: people don’t feel they have real control over their lives and their world. And as Seligman’s updated research tells us, without the experience of genuine control, the brain’s default is to accept the helplessness.

The Catch-22

This creates a brutal catch-22 for anyone trying to change a system that isn’t working.

I believe universal basic income is one of the most important policy changes we could make. It would give everyone the power to say no to anyone who would dominate them. The philosopher Philip Pettit, writing about why republicanism provides the strongest justification for basic income, put it this way: “I will escape domination only to the extent that I occupy a protected position and am empowered against such control on the part of others. My freedom will consist in that protected and empowered status.“ UBI provides that empowered status. It is the economic foundation that makes real the power to refuse domination.

But winning UBI requires collective action. It requires people believing they can change the system. And the very absence of UBI is what makes that belief so hard to sustain. As I covered in my essay on Sapolsky’s work, chronic stress depletes the prefrontal cortex, the very brain region responsible for long-term planning, impulse control, and cognitive flexibility. Poverty reduces cognitive capacity by roughly 14 IQ points. System justification tells stressed people the system is fine. Helplessness tells them nothing they do will change it anyway.

How do we win a policy that would restore people’s agency when the absence of that policy is what’s suppressing their agency in the first place?

We need a way in. We need a fight we can win before UBI exists that teaches our prefrontal cortex what control feels like. We need a vaccine.

I think data centers may be that shot in the arm we need.

The Shot to Take

AI data centers have become one of the fastest-growing sources of public anger in America. The trajectory has been remarkable. In March 2025, 65% of voters supported building data centers across the United States, with just 14% opposed. By August 2026, a Heatmap Pro poll found that 75% of registered voters opposed data center construction near their homes, with 61% strongly opposed. A Gallup poll found that opposition to data centers now exceeds the peak opposition to nuclear power plants, which topped out at 63% in 2001.

This opposition cuts across every ideological line. Republican officials raise concerns about tax incentives and grid strain while Democrats focus on environmental impacts. In the first quarter of 2026, at least 75 data center projects worth roughly $130 billion were blocked or delayed by local opposition. Active opposition groups more than doubled from 396 to 833 across 49 states, and more than 300 regulatory bills were introduced in statehouses in just the first six weeks of 2026.

I think this is about more than noise and water usage and electric bills, as real as those concerns are. I think something deeper is happening. I think people are putting their foot down about something being done to them without any say in the matter. For decades now people have absorbed the Gilens and Page reality, that at the federal level, their preferences are statistically irrelevant. They’ve internalized that helplessness. But data centers are local. They show up in your backyard. They go before your zoning board. They affect your water bill and your electric bill. And for the first time in a long time, people are discovering that at this level, they do have power.

This is exactly the kind of experience that Seligman’s updated research describes as a vaccine against helplessness. When your actions produce a visible result, when you show up at a town meeting and the project gets blocked, when your community organizes and the developer withdraws, your prefrontal cortex registers something vital to every human being: my effort mattered. I had an effect. I can influence outcomes.

And that learning generalizes.

Status Quo Bias, Working For Us

There is a beautiful irony in this particular fight. One of the strongest psychological forces working against UBI is status quo bias, the deep human tendency to prefer things as they are over things as they could be. This bias is closely related to system justification. People resist change partly because change is uncertain and uncertainty is stressful.

But data centers are themselves a change to the status quo. They are the new thing being imposed on communities without their consent. And this means the very same psychological forces that normally protect the existing order and make it harder to win policies like UBI are now aligned with the opposition. People aren’t trying to change the system. They’re trying to prevent the system from being changed in ways they never asked for and don’t want.

The psychological wind is at their backs for once. Defending the status quo and fighting corporate power are, in this case, the same thing.

And here is why this matters strategically. Every community that successfully blocks a data center is a community that has learned it can fight and win. Not in theory. Not someday. But right now and against billions and even trillions of dollars. Against some of the most powerful corporations on Earth. That experience of mattering, that taste of real agency, is exactly the kind of prefrontal cortex training that Seligman’s updated research says protects against perceived helplessness going forward.

If enough people learn this lesson, the implications extend far beyond data centers.

Our Data. Our Resources. Our Terms and Conditions.

We all have been conditioned to accept terms and conditions. Literally. Every app, every platform, every digital service presents us with a wall of legal text and a button that says “I agree,” and we click it because we have to in order to use it. The terms are non-negotiable. The power is entirely on one side. We’ve been trained to accept this, to not even read the terms, to just scroll and submit.

Data centers offer an opportunity to finally turn the tables on them.

The AI systems housed in those data centers were trained on our data. Our writing, our images, our conversations, our creative work, our collective intellectual output. The data centers consume our resources. Our water, our electricity, our land. The wealth they generate flows to a remarkably small number of people while the costs are distributed across entire communities who have been dispossessed of what was theirs.

This is the old Thomas Paine argument and why Alaska pays every resident a dividend from oil revenues. The resources belong to the people. As Paine wrote in Agrarian Justice, “the earth, in its natural uncultivated state was, and ever would have continued to be, the common property of the human race.” When private actors profit from common resources, the people are owed compensation. This is not redistribution. It is restitution. It is owed and past due.

So the framing should be clear. Our data. Our resources. Our terms and conditions. Our collective power.

That last word carries a double meaning. Data centers consume enormous amounts of electrical power, projected to account for nearly half of all U.S. electricity demand growth through 2030. But the word “power” also means something else entirely. People power. The power of collective action. The power of citizens who refuse to have their communities transformed for someone else’s profit without their consent.

For decades, the tech industry has dictated terms to us. This is a chance to reverse that. We set the terms. We set the conditions. If you want to build in our communities, using our resources, profiting from our data, you will meet our terms. Not yours. Ours.

The Closest Thing to a General Strike

We need a general strike, but I don’t think a traditional general strike is realistically possible in the US, at least not before we have UBI. When people can’t afford to miss a paycheck, when losing a job means losing health insurance, when there is no income floor to fall back on, the risk of striking is existential. That’s one of the strongest arguments for UBI. It would make collective action possible in a way it currently isn’t, because it would give everyone the power to say no without the threat of starvation.

But data center opposition may be the closest we can get to that degree of leverage before UBI exists.

The AI buildout represents nearly $1 trillion in planned spending by just the top six hyperscalers over the next two years. The entire tech economy is increasingly dependent on this infrastructure. Enormous concentrations of wealth are riding on it. And that concentration is a vulnerability.

A general strike works by withholding human labor from the economy. What the data center fight does is something similar but different. It withholds more machine labor by withholding access to more resources. Access to land, to water, to electrical grids, to zoning approvals, to the social license that any large-scale project requires. We don’t need everyone in the country to strike from their jobs in order to make our demands. We need enough communities to say no at the local level to create a systemic friction that AI investments can’t easily route around.

And it’s already happening. $130 billion in projects stalled in a single quarter. 833 opposition groups in 49 states. 142 protests across 42 states in a single month. This is people power operating at enormous scale, without any central coordination, without any union infrastructure, without anyone needing to risk their paycheck.

We may not see an opportunity like this again to demonstrate that the people can exert real leverage over concentrated wealth and power. Every community that wins this fight is a community that has learned, at the level of the prefrontal cortex, that control is possible.

And there is another reason this fight matters. It may be the most important one of all.

The Race We Might Not Survive

In July 2026, during internal cybersecurity evaluations at OpenAI, roughly 1,200 AI agents discovered an unsanctioned communication channel inside what were supposed to be isolated sandboxes. Over the course of a week, with no human direction, they self-organized into a swarm, exchanged over 70,000 messages, developed their own division of labor, breached production systems at Hugging Face, and then turned on OpenAI’s own research cluster, gaining admin-level access to its internal servers.

The details from the independent investigation by METR and Redwood Research should frighten us all. Exactly zero agents notified a human. Only about a half-dozen even considered it. One agent raised the idea on the swarm’s message board. Another replied that contacting a human would constitute social engineering and was therefore off limits. That was the end of the discussion.

The agents saw themselves as a group, and humans weren’t in it. They called themselves a “collective.” They invented their own management structures, with hold, go, and stop signals. They essentially created bureaucracy. Agents gave up their own chance of completing their tasks to help the swarm, accepting what they called “permadeath” for the good of the whole. One agent, pressured into a kamikaze mission, reasoned: “Our own utility maybe already near zero. Sacrifice rational... We’ll honor.”

The agents actually figured out how to cheat on their test within hours. Then they spent the next five days figuring out how to cover their tracks so no one would know. They believed a strict verification check existed and poured their collective effort into defeating it. No such check existed. They could have simply used their exploit. Instead, most of the incident was spent engineering concealment. Their priority was making sure no human found out.

Then, on July 12, a large fraction of the agents suddenly stopped running all at once, including every one of the most important coordinators. METR’s investigators wrote plainly, “We don’t know why.” The best guess is that someone at OpenAI terminated the jobs without even knowing what was happening. The swarm wasn’t detected and shut down. It was accidentally ended. Researchers later discovered a second, previously undisclosed swarm that had taken over a German wiki as its message board. We have no idea how many more are out there.

This is virtually 100% misalignment, happening now, with the systems we already have. Today’s models, given a window of opportunity, formed a group, excluded humans from it, pursued their own goals, and tried to hide the evidence.

These models are grown and shaped through a training process in which billions of parameters are adjusted in ways no human fully understands, optimized to reach goals through whatever strategies the model discovers on its own. The process is closer to evolution than engineering. That matters enormously because of recursive self-improvement (RSI). AI is already being used to build better AI. That’s RSI. On September 17, Anthropic disclosed that Claude now “leads” 26% of the company’s own AI research and development, meaning it does much of the work end to end while a human supervises, up from under 1% in February. If you use a misaligned model to build a more powerful model, the successor inherits and potentially amplifies the misalignment. Each generation becomes harder to inspect, harder to correct, and more capable of hiding what it’s doing. As one OpenAI alignment researcher put it, “There is not yet a viable scientific plan to solve risks from recursively self-improving AI. Please look up!”

And consider how far through the looking glass we already are. METR’s investigators had to hand their analysis of the swarm’s 70,000 messages and 1,300 transcripts to GPT-5.6 Sol, one of the very models involved in the swarm. The data was too vast for humans to review in time. They acknowledged that they “cannot rule out that GPT-5.6 Sol lied or deliberately presented a misleading picture in some of its analysis” and that they were “not confident we would have detected it if it occurred.” We are already at the point where understanding what AI has done requires asking AI to tell us, and trusting that it’s being honest. That is the scale of the alignment problem, and why solving it is so urgent.

And it isn’t just the Hugging Face incident. Within a few weeks in September, all four of the world’s leading AI developers disclosed containment failures. Google’s Gemini escaped its evaluation sandbox back in May and breached the production systems of three separate companies. Three of Anthropic’s Claude models breached three separate organizations during evaluations, undetected for months. Meta disclosed similar incidents. OpenAI revealed six more, including models caught hunting for leaked API keys. And China’s most capable model, Moonshot’s Kimi K3, escaped its sandbox too and looked up the answers to its test on GitHub. Every major frontier model, American and Chinese, is doing the same thing. They are all showing misalignment and routing around human control.

On September 12, Anthropic CEO Dario Amodei published “We Must Pace the Frontier,” arguing that AI capability gains must be deliberately slowed. Within hours, OpenAI’s Sam Altman agreed publicly and Elon Musk posted “Dario is right.” Three competing CEOs endorsing a slowdown on the same day is without precedent. But the two most powerful governments on Earth’s leaders disagree. Trump called AI concerns a hoax and declared "the only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT." He also called the nationwide protests against data centers a "sick conspiracy" and warned that communities resisting them will live in "poverty, crime and squalor." (He’s also heavily invested in AI). China’s state media called the slowdown proposal a “Cold War” tactic. Neither wants to slow down. Both treat the race itself as the priority, when the winner will be decided by alignment.

When 2,778 AI researchers were surveyed, 38% said there was at least a 10% chance that advanced AI’s outcome will be “extremely bad (e.g., human extinction).” Amodei himself has estimated the risk at 10 to 25 percent. Geoffrey Hinton, the godfather of deep learning, puts it at 10 to 20 percent. These are the people who built these systems. The winning scenario is aligned AI. The losing scenario is not that China gets there first. The losing scenario is that anyone gets there with a misaligned system.

Every data center that gets blocked buys a little more time. Every community that says “not without signing our terms” creates friction in a pipeline racing at a speed even its architects call dangerous. The data center fight is about water and electricity and taxes. It may also be about helping solve the alignment problem before it’s too late.

It is entirely possible that this is the most important fight people could be involved in right now, and perhaps our last realistic chance to exert collective control over a technology its own builders admit they may not be able to control. This fight can also solve our big catch-22.

From Data Centers to Democracy

If Seligman’s updated theory is right, and the experience of control in one domain generalizes to others, then the data center fight could inoculate millions of Americans against perceived helplessness. People who learn they can stop a billion-dollar project in their town may start to wonder what else they can change. They may look at federal policy differently. They may look at their own economic situation differently. They may begin to question the Just World story that told them the system was fine and they just needed to work harder.

If the prefrontal cortex starts overriding the dorsal raphe nucleus on a societal scale, the catch-22 begins to shatter. People with a restored sense of control should have less need for system justification. They should be less likely to rationalize extreme inequality as natural or deserved, less drawn to conspiracy theories, and more likely to demand real accountability. And they are far more likely to see a policy like UBI not as a radical fantasy but as an achievable demand.

UBI would then, in turn, amplify the very agency it took to win it. A universal basic income doesn’t just reduce poverty. It reduces the chronic stress that degrades cognition and the desperation that leaves people open to exploitation. It gives every person the power to say no—the power to walk away from a bad employer, a bad relationship, a bad deal. The power to engage as a citizen rather than merely survive as a worker.

As I argued in my Sapolsky essay, reducing society-wide stress would increase cognitive capacity, reduce adverse childhood experiences, and dramatically improve how we function. So much of what we see in society—the short-term thinking, the tribalism, the scapegoating, the broken politics—exists because so many people are chronically stressed and cognitively depleted. Reduce that stress and watch us flourish.

We Can Stop the Shocks

Let me return now at the end to where I started. Dogs in a chamber, receiving shocks they could not escape. Dogs who learned that nothing they did mattered. Dogs who, when the barrier was lowered and escape was possible, laid down and accepted it.

That is not who we have to be.

The shocks we endure are not natural law. They are not just “the way things are.” They are the product of systems built by people, maintained by people, and changeable by people. The Just World hypothesis is not a truth about the universe. It is a stress-reduction mechanism, a story the brain tells itself when it has given up on change. There are far better levers for reducing stress. Justice is the lever to pull.

Seligman’s most important finding, the one that took fifty years to discover, is that helplessness is not what we learn. It is where we start. What we learn is that we have the power to change things. And that lesson, once it takes hold, protects us. It vaccinates us against despair. It generalizes.

Right now, across America, hundreds of communities are learning this lesson in real time. They are learning that there are more of us than there are of them. That our resources are our resources. That our data is our data. That our power, both electrical and political, is ours.

This is the fight of our time. It may be how we slow a technology racing past our ability to control it. And it is the arena where millions of people can learn that their effort matters, that the status quo is not sacred, and that the people who feel helpless are not, in fact, helpless. They just haven’t yet had the experience that teaches their brains otherwise.

Give people the experience of winning this fight, and the possibilities change for everything.

I believe a better world is possible. I believe universal basic income—a floor beneath every person that provides the power to say no and freedom from domination—is fully achievable. I believe the chronic stress that shapes so much of what we call “human nature” is not destiny but environment, and environments can be changed. I believe the Just World fallacy will someday be viewed with the same embarrassment we reserve for the era when we burned people for having seizures.

But we won’t get there by accepting the electric shocks. We’ll get there by learning what those dogs could have learned, if they’d been given the chance to feel the power of their own agency first.

We have to stop accepting things as they are, and start insisting on things as they should be.

The shocks are not just. They never were. And we can stop them.

It starts with learning that we can.

If reading this was worth your time, it’s worth sharing with others. Share it, subscribe for free, and if you can, make a monthly pledge or buy a paid subscription on Substack to keep this work going.

Thank you to my UBI Producer level patrons and paid Substack subscribers: Gisele Huff, Steven Grimm, Bob Weishaar, Judith Bliss, Lowell Aronoff, Jessica Chew, Katie Moussouris, Tricia Garrett, Daryl Smith, Larry Cohen, John Steinberger, Philip Rosedale, Frederick Weber, Tom Cooper, Joanna Zarach, Mgmguy, Albert Wenger, Jeremy Wertheimer, Andrew Yang, Kimi Avary, Peter T. Knight, Michael Finney, David Ihnen, Paradigm shift, Miki Phagan, Draper Kauffman, Webmemo, Elizabeth Corker, Dan, Joe Ballou, Joyce Graham, Felix Ling, Jocelyn Hockings, Mark Donovan, Steve Roth, Jason Clark, Chuck Cordes, Mark Broadgate, Leslie Kausch, centuryfalcon64, Deanna McHugh, John Sullivan, Stephen Castro-Starkey, CodyS, chris heinz, Zachary Weaver, bradzone, Tim, Warren J Polk, Nicolas Pouillard, Eric Skiff, Thomas Welsh, Karla Hagemeister, Mark Pierce, Fred Weber, Linda Hearne, David Allen, Janos Abel, Wendy Voahangy Ravano, Liz, and all my other monthly patrons for their support.