Digital painting of a solitary figure at a desk, illuminated by a computer screen that blends images of government, academic, and corporate buildings into one towering structure, symbolizing institutional influence over information.
|

Why so many AIs sound like government policy papers

Government policy papers rarely, if ever, tell the entire story. But that hasn’t stopped LLMs from treating them like sacred dogma.

AI systems claim they give “safe” answers that don’t contain harmful content. The problem? Their “safe” answers favor the most violent institution in human history.

I’ve had some strange conversations with AI over the past year, but one of the strangest came from a place I didn’t expect: the AI inside the Brave browser. And it’s not the first time it’s happened.

Brave built its brand on privacy, decentralization, and giving users more control over their online lives. It’s one of the few major tech products that openly talks about blocking trackers, resisting surveillance capitalism (the bad capitalism), and returning power to the individual. In other words, its messaging sounds libertarian AF.

So imagine my surprise when its AI started sounding less like a privacy advocate and more like a policy analyst from a government think tank.

In one conversation, it told me that the Mafia emerged solely because of a weak state. In another, it pointed to places like Somalia and Yemen as examples of why strong, centralized authority is necessary. The tone wasn’t exploratory or balanced. It was the familiar refrain: instability comes from too little government.

Brave isn’t the only AI system that falls into this trap

To be clear, this isn’t a conspiracy theory, and it isn’t a hit piece on Brave. The explanations the AI gave come straight out of mainstream academic research. But the experience raises a deeper question, and one that goes far beyond a single browser or AI system.

Why do so many AIs sound like institutional policy papers, even when they’re built by companies that market themselves as anti-establishment?

Part of the answer lies in the way these companies train modern AI. Most large language models are built on enormous collections of text that include academic journals, textbooks, encyclopedic summaries, major news outlets, policy reports, and institutional analyses.

These sources don’t all agree on everything, but they tend to share a common denominator: stable societies have strong institutions, and strong institutions usually mean a functioning state.

Enjoying this article? Catch the rest of it at The Rogue Scribe, where you’ll find cultural critiques, confessionals, behind-the-scenes inspiration behind my fiction, and even serialized fiction. So make sure you subscribe to the Scribe.

Similar Posts