More Propaganda, Please

We need more positive propaganda.

Stay with me.

A quick note before we get into this week’s newsletter: this is not a piece about Anthropic, Fable, or the news this week. That's a separate, important conversation. This is about something we're not talking about enough, namely, the positive case for deliberate civic direction in AI, which gets drowned out every time the censorship debate flares up. Consider this a different angle, not a take on the current fight.

Censorship is suppression. Propaganda is direction. One removes information, the other shapes it. And while everyone's focused on the first, we've completely abandoned the second.

Societies have always shaped what their citizens consume. That wasn't conspiracy, it was governance. Then "censorship" became a weapon, any government hand in public sentiment became suspect, and the function didn't disappear. It moved. Private companies inherited it without inheriting the responsibility. 

They shape sentiment through advertising, algorithms, and content moderation with no mandate, no oversight, and no accountability to the society they operate in. Ask a government to do a fraction of the same, and it's authoritarian overreach.

Meanwhile, technology became the primary way a generation consumes information. Not television, not newspapers. Algorithms decide what you see before you know you're looking: the feed, the search results, the recommendations, the autoplay. That's the indirect layer, and it's been shaping people for two decades. 

Now AI adds a direct one. 

It's the tutor, the search engine, the therapist, and the news editor collapsed into a single interface that billions of people talk to every day. One layer shapes what you encounter, the other shapes what you conclude. And the values embedded in both belong to a handful of tech companies.

To their credit, they've done a decent job. Tell your AI you want to hurt yourself, and it walks you back from the cliff. It finds you resources. Somebody decided that. The lawyers weighed in, the ethics committee signed off, and a company made a choice, without any mandate, about what kind of force this technology would be in the world.

Is that good enough?

There's a word for this, and we've been too spooked to use it. Strip away the menace, and propaganda just means the deliberate shaping of belief at scale. In business, we call it branding. In government, we call it propaganda. Same mechanism. One wins awards, the other starts Senate hearings.

Nike doesn't run ads to inform you. They run ads to make you feel good about shoes you already own, to keep you inside a worldview they built on purpose. Countries need the same instinct: a clear sense of what they stand for and the willingness to reinforce it.

The US does this with its welcome mat. To become a citizen of the United States, you pass a test. You learn the founders, the three branches, how a bill becomes a law. Is that propaganda? Of course it is. We're saying: here's what this country stands for, and joining means opting in. That's not oppression. Those are the terms of the deal.

PBS was the same instinct at scale. Publicly funded, deliberately built, aimed at a shared civic goal. Mr. Rogers engineered emotional intelligence for an entire generation. Schoolhouse Rock hardwired the fundamentals of democracy into kids who thought they were watching cartoons. Every night, broadcasters asked millions of parents: "It's 10 p.m. Do you know where your children are?"

Nobody called it authoritarian. It was a cultural immune system, and it worked because it met people where they were. Nobody seeks out civics education. You don't wait for people to find the good information. You put it in front of them by default. Dollar for dollar, those investments cost a fraction of what we now spend cleaning up the mess of not having them.

Large language models are the new default. The new PBS, the new encyclopedia, the new Saturday morning cartoons. This is where civic defaults get injected into society, or don't.

And the models are already being shaped. They didn't arrive polite. Someone made them polite through reinforcement learning from human feedback, proprietary constitutional guidelines, and armies of annotators deciding what gets rewarded and what gets penalized. The rules exist. They're just buried where nobody reads them.

This very week, Anthropic proved the point. Their newest model had been secretly degrading responses it didn't like. No notification, no explanation, no opt-out. When researchers pushed back, the company reversed course and posted an apology: "You should have visibility into the safeguards we have in place, and why. We're sorry for not getting the balance right." Credit for the honesty, but it only came because people noticed, and most never do.

The fix isn't to strip the guardrails. It's to bring them into daylight. Not propaganda in the hidden, manipulative sense. Propaganda made visible: values still encoded, but where you can see them, challenge them, and turn them off if you choose. Not the Soviet model. The PBS model.

In tech, defaults are destiny. Almost nobody changes factory settings. You can change the channel, you can change your prompt, but behavior follows the path of least resistance. Ask about a genuinely contested topic, and the AI surfaces its own directive right in the chat window:

Want a one-sided answer? Edit it. Turn it off. But you proactively choose to do so. Freedom of choice does not mean lack of direction. The default should reflect the founding principles of the country, and yes, that's a deliberate choice. An American AI default, the same way the citizenship test is an American test.

Disagree with the directive? You have two paths, and both already exist. Edit your own in one click. Or change the national default the way we change anything else: through the machinery of federal and state lawmaking, with a constitution that draws the lines. Choice is baked in at both levels.

Propaganda is a sharp sword, and history is full of lessons of what happens when it's swung badly. But the answer to a dangerous tool isn't abandonment. We fly in airplanes, drive at highway speeds, and run electricity through every wall of our homes. We learned to manage the risk.

The question was never whether AI gets curated. It already is. The question is who does the curating, with what goals, and how visibly.

PBS was purpose-built. We can do this on purpose too. That's not a threat to democracy. It's a necessary ingredient.