When Bad Policy Hides in Plain Sight: Tariffs, AI, and the Politics of Not Being Told No

May 01, 2026

10 min read

8 views

84%

The strange coincidence nobody should ignore

What do a tariff formula, a neo-Nazi account deleting its posts, and a White House full of courtiers have in common? At first glance, almost nothing. But together they point to a deeper modern problem: institutions now fail not only by making bad decisions, but by making bad decisions that can be justified, erased, or automated after the fact.

That is the real story underneath the tariff chaos. The obvious version is political theater: a president dislikes trade deficits, slaps tariffs on imports, and markets panic. But the more interesting version is that a policy can be both obviously reckless and yet strangely defensible to the people inside the machine, because the machine rewards the wrong kinds of evidence. A bad formula can look sophisticated. A bad idea can be dressed up as national strategy. A bad actor can delete the record and hope the trail disappears.

The web archive keeps the receipts. The market keeps the score. The only question is whether the system learns before the damage compounds.

Modern power is not just about deciding. It is about controlling the narrative trail that makes the decision look inevitable.

The new politics of plausible nonsense

There is a particular kind of policy error that is uniquely dangerous in the age of AI and social media: plausible nonsense. It is not obviously fake. It is not mathematically impossible. It is merely disconnected from how the world actually works.

That is why the tariff calculation matters. If you reduce a country’s trade relationship to a bilateral deficit, divide by imports, and call the result a tariff equivalent, you can create a spreadsheet that looks rigorous enough to pass in a meeting. You can even get a language model to generate a similar method. But the existence of a formula is not the same thing as the existence of a theory.

This is where policy becomes dangerous. A real economic argument has to survive contact with messy reality: supply chains, rules of origin, investment timing, retaliation, exchange rates, and the fact that people respond to incentives in non linear ways. A fake argument only has to survive a slide deck.

Consider the practical absurdity of country specific tariffs. If goods from the European Union are taxed one way and goods from Britain another, then a ship crossing the English Channel for five minutes suddenly raises a question that can absorb entire bureaucracies: where did the product really come from? That is not a side issue. It is the hidden cost of turning trade policy into a moral scoreboard. Once you do that, the economy no longer runs on commerce. It runs on paperwork.

The broader lesson is bigger than tariffs. AI systems are extremely good at producing the grammar of expertise and extremely bad at guaranteeing the substance of expertise. They will happily synthesize the internet’s most repeatable patterns, including the bad ones. If the output sounds like something a serious person might say, the temptation is to mistake coherence for correctness.

That is not just an AI problem. It is a governance problem. In a political culture where people are rewarded for sounding certain, the most dangerous advice is the advice that sounds neat.

The courtier state and the danger of nobody saying no

There is a reason this policy style feels so unstable. It is not only that the president likes tariffs. It is that the surrounding ecosystem has learned how to speak to him.

That is the core of the courtier problem. When power becomes personalized, the people who survive are not the people who tell the truth. They are the people who intuit what the leader already wants and then retrofit justification afterward. The result is not strategy in the classic sense. It is a loyalty cascade.

This matters because bad policy is often explained as if it came from a deep ideological blueprint. Sometimes it does. But sometimes the real engine is simpler and more unsettling: the leader has a crude instinct, and everyone around him searches for a theory that flatters it.

That helps explain why contradictory ideas can coexist so comfortably. One faction wants America to be stronger and more feared. Another wants America to stop “paying” for the global order. One faction wants the dollar to remain dominant. Another blames that dominance for deindustrialization. One faction wants to deglobalize. Another wants leverage over allies. The contradictions are not a bug. They are a feature of court politics. Different people are free to project their preferred doctrine onto the same leader, because the leader is the doctrine.

This is why waiting for “Republican grownups” to intervene is so often a losing bet. In systems like this, dissent is not merely unpopular. It is structurally self defeating. To oppose the ruler is to risk exile from the coalition, and in a personalized movement, exile is often career death.

A courtier system does not need everyone to believe the same thing. It only needs everyone to fear saying the true thing too loudly.

That is how bad ideas acquire a false aura of inevitability.

Why trade deficits became a moral obsession

To understand the emotional force behind tariffs, you have to understand the symbolism of the trade deficit. Economists can explain that deficits are not the same thing as failure, and that in the U.S. case they often reflect the country’s attractiveness to global capital, its strong productivity growth, and the fact that the world wants to hold its assets.

But symbols travel differently than data.

For many voters, a trade deficit feels like a humiliation story. Someone else is selling more to us than we sell to them. That sounds like we are being taken advantage of. It feels like a scoreboard, and people naturally want the score fixed. Tariffs then promise not just economic repair but dignity. They offer the emotional satisfaction of punishment.

This is why trade politics is so much more volatile than trade economics. The economist sees a system of interdependence shaped by comparative advantage, capital flows, automation, demographics, and productivity. The nationalist sees a story of winners, losers, and people “getting away with it.” These are not merely different conclusions. They are different moral grammars.

The hard truth is that tariffs are a blunt tool for what many people want them to do. They do not reliably re industrialize an economy. Manufacturing employment has fallen mostly because of automation and productivity growth, not because the country failed to tariff enough. Even if trade deficits vanished, the share of manufacturing jobs would not snap back to some mid century level. The economy has changed too deeply for that fantasy.

A useful mental model is this: trade deficits are a shadow, not always the substance. If you stare at the shadow long enough, you may miss the object casting it. The object is often something less dramatic and more durable, such as technology, skills, bargaining power, and industrial policy.

That is why “tariffs will bring jobs back” is such a seductive but shallow promise. It mistakes visible borders for invisible forces.

The real industrial policy is less cinematic and more boring

If the goal is national resilience, the relevant question is not whether to punish imports in the abstract. It is which production should be located where, under what conditions, and for what strategic purpose.

There is a serious national security argument for reshoring or friend shoring certain goods. Semiconductors are the obvious example. If a country depends on a distant supplier for components that can paralyze defense, medicine, or advanced manufacturing, that is a vulnerability worth addressing. But that is not the same as taxing sneakers from Bangladesh or furniture from Vietnam into oblivion.

The difference matters because security policy and consumer punishment are not the same thing.

A smart industrial policy is targeted, patient, and specific. It asks where resilience matters, where redundancy is worth paying for, and where alliances can reduce risk. That is why the logic of the CHIPS Act makes more sense than a blanket tariff regime. If you want secure supply chains, you subsidize, coordinate, and plan. You do not randomly raise the price of everything and hope patriotism fills the gap.

This is also where the fetish for simple formula becomes dangerous. Real industrial policy is not tidy. It involves trade offs. It means accepting that some sectors need support, some need friends, and some should simply be left alone. It does not produce a single number that can be cut in half and printed on a chart.

The temptation to turn complexity into a formula is understandable. It gives leaders the feeling of mastery. But the economy is not a moral ledger. It is a living system. When you hit it with blunt instruments, the pain often lands somewhere else than you expected.

A good rule: if a policy promises both national revival and spreadsheet simplicity, be suspicious.

AI is not the villain. It is the amplifier

It is too easy to blame AI for all of this, as if the machine invented the mistake. That is too generous to the humans involved and too forgiving of the political environment.

Language models do something subtler and more dangerous. They compress the internet’s surface layer of argument into something fluent and immediate. If the internet contains simplistic trade heuristics, outdated nationalism, conspiratorial thinking, and cherry picked economic claims, the model can produce all of that in a voice that sounds polished. It is not a thinker. It is a mirror with autocomplete.

That makes AI less like a mastermind and more like a bullshit accelerator. It can turn half formed intuitions into pseudo policy faster than a room full of hesitant experts can stop them. And if the people in power already prefer flattery to correction, the machine becomes a perfect accomplice.

This creates a governance problem that is easy to miss. Traditional accountability assumed bad ideas would be visibly authored by identifiable humans. But now bad ideas can appear as if they came from nowhere: from a model, from a memo, from a fog of consensus, from “what the data says.” The origin story becomes blurry just as the consequences become concrete.

That is why the deleted neo Nazi posts and the archived record are part of the same story. In both cases, the fight is over traceability. Who said what, when, and under what incentives? What was erased? What was preserved? What was laundered into respectable language?

The archive matters because memory is now a policy variable.

Key Takeaways

Treat fluent output as a starting point, not evidence. Whether it comes from a person, a memo, or an AI model, coherence is not correctness.
Separate emotional appeal from policy mechanism. Tariffs can satisfy a punishment instinct without solving the problem they claim to address.
Ask what problem a policy is really solving. If the goal is resilience, targeted industrial policy usually makes more sense than broad retaliation.
Watch for courtier dynamics. When leaders reward agreement over expertise, the bad ideas around them become more extreme and less accountable.
Look for the hidden force behind the shadow. Trade deficits, market anxiety, and deindustrialization are often symptoms of deeper changes like automation, capital flows, and bargaining power.

The deeper lesson: systems fail when they confuse legibility with truth

The most unsettling connection across all of this is not tariffs, AI, or online extremism by themselves. It is the way modern systems increasingly reward whatever can be made legible fastest.

A tariff formula is legible. A chatbot answer is legible. A deleted post with an archived snapshot is legible after the fact. A leader who says he alone understands the deal is legible to supporters who want certainty more than accuracy. But the world itself is not legible in that way. It is messy, path dependent, and stubbornly resistant to slogans.

That is why the current moment feels so disorienting. We are surrounded by tools that can produce the appearance of order faster than institutions can verify whether order exists. We can now fabricate certainty at industrial scale.

The challenge, then, is not merely to oppose bad policy. It is to rebuild a culture that prefers explanation over performance, evidence over vibes, and traceable truth over convenient narrative.

Because once a country starts governing by what sounds right in the room, what can be written into a prompt, and what can be erased from a feed, it is not just making tariff mistakes. It is training itself to accept unreality as competence.

And that is a much bigger trade shock than any tariff schedule.