Big Tech's Censorship Machine Has a First Amendment Problem

Here's the Thing Nobody Wants to Admit

I helped build content moderation systems. I know exactly how they work, because I wrote some of the code that runs them. And I need to tell you something that the "private companies can do what they want" crowd doesn't want you to hear: when those private companies are taking direction from government officials about what to suppress, it's not content moderation anymore. It's censorship. Full stop.

The Twitter Files — released in late 2022 and early 2023 — documented something that technologists already knew and civil libertarians feared: the FBI, DHS, and other government agencies maintained regular communication channels with social media platforms, flagging specific posts and accounts for suppression. Not suggesting. Flagging. With the implicit understanding that noncompliance carried regulatory consequences.

The Technical Architecture of Censorship

Let me explain how this actually works, because the mainstream coverage treats it as abstract policy when it's actually engineering.

Every major social media platform maintains a system called a "trust and safety" pipeline. Content enters the pipeline through automated detection (machine learning classifiers), user reports, or — here's the critical part — "trusted flaggers." Trusted flaggers are external entities that can submit content directly into the review pipeline with elevated priority. Government agencies were trusted flaggers.

Spoiler alert: the code doesn't distinguish between a "recommendation" from the FBI and an "order" from the FBI. It processes both the same way. The content gets suppressed. The user gets a strike. The algorithm downgrades future content from that account. The code doesn't lie. But the press release calls it "partnership" instead of "censorship."

Think about that for a second. The government can't censor your speech directly — the First Amendment prohibits it. But it can call your social media platform, "flag" your post as "misinformation," and watch it disappear within hours. Same result. No warrant. No due process. Just a phone call.

The Legal Reckoning

The Murthy v. Missouri case — which reached the Supreme Court in 2024 — was supposed to resolve this. The Fifth Circuit ruled that the government's communication with platforms constituted coercion, violating the First Amendment. The Supreme Court ducked the issue on standing grounds.

The legal question remains open. But the technological reality doesn't: the infrastructure for government-directed censorship exists. It's been built. It's been used. And the only thing preventing its expansion is the political cost of getting caught — which, as the Twitter Files demonstrated, is approximately zero.

What Needs to Happen

Decentralize everything. Social media. Search. Payment processing. DNS. Every chokepoint that can be controlled by a single entity will eventually be controlled by someone with bad incentives. The architecture of the internet was designed to be decentralized. We centralized it for convenience. Now we're paying the price.

Until we rebuild the digital public square on infrastructure that no single company or government can control, "free speech" will remain a constitutional guarantee with a terms-of-service asterisk.

They're not bugs — they're features.