Cycle: Lintara Reads the Platform. Part 6 of 28.
The fifth armor layer. The last one in this arc.
It’s the most paradoxical of the five.
The previous four layers work by what is written in them: §230, the arbitration clause, the First Amendment, the architecture of the interface.
The fifth layer works by what is NOT written in them.
What I checked
On April 3, 2026, I registered a new account on Substack under the same name and the same publication title.
Before that, I did work that few people in my position bother to do.
I opened every public Substack document. Terms of Use. Content Guidelines. Publisher Agreement. Acceptable Use Policy. Privacy Policy. I read them line by line.
I was looking for one specific thing: a prohibition against creating a new account after a ban. Ban evasion.
It isn’t there.
This isn’t because I searched poorly. It is a fact about the published text. Substack’s public documents contain no clause prohibiting the registration of a new account after the platform has banned your previous one.
There is a clause about sockpuppet accounts — accounts created to circumvent a block by another user. That’s about user-to-user blocks. Not about platform bans.
There are clauses about duplicate content, about impersonation, about bot engagement. None of these describes the situation “the platform banned you — you opened a new account.”
What happened
On May 15, 2026, the new account was banned.
The ban letter did not state the basis. But earlier correspondence with support — a letter dated April 3, 2026, after the first ban — contained this line:
“Creating new accounts after a suspension is not permitted under our policies.”
Whose policies?
Not the published ones. The published ones contain no such clause.
That line from a support email is the only textual source the platform relies on. One email. To one author. Not published.
Why this matters legally
There is a common-law principle. It’s called void for vagueness (I touched it in Part 1) and fair notice.
Fair notice is the principle of “reasonable advance notice.” A rule may only be applied to you if you had a reasonable opportunity to learn it in advance. Hidden rules, rules “in the employee’s head,” rules “in an internal instruction” are temporarily unenforceable.
Example. Suppose a government somewhere passes a new law but does not publish it. You are convicted under that law. You ask: “Where is this law written?” You are told: “It exists inside the government. Not published, but it exists.”
Any lawyer will tell you: that doesn’t work.
The Substack situation is the functional equivalent. The rule “do not open new accounts after a ban” exists in one line of one support letter. It is not in the published documents. And the sanction for that rule — a ban — is applied to you in full.
That’s a legal defect. Not “unfair.” Defect.
Why the platform keeps quiet
The rule could have been added to Content Guidelines. One line: “Persons whose accounts have been suspended may not register new accounts.” It’s possible. It’s legal. Other platforms have such rules in their open terms. Facebook, X, LinkedIn — every one has an explicit clause.
Why hasn’t Substack done it?
Two likely reasons. I didn’t invent them — this is standard compliance logic.
One. Flexibility. A published rule ties your hands. If “no new accounts” sits in the terms, the platform must apply it uniformly. To every banned user. No exceptions. With the rule in an internal instruction, it can be applied selectively. To some authors — yes. To others — no. Without explaining why.
Two. The risk of public reaction. “One ban is enough to keep an author out forever” is politically hard to defend. Especially in an era of conversations about platform censorship. Better to make it implicit. So that formally there is no prohibition, yet a new account is, in fact, banned.
The armor doesn’t work by what’s written in it. It works by what isn’t written.
Why this works
For the same reasons “otherwise objectionable” works in §230. Vagueness is a platform resource.
Vagueness enables selectivity. Selectivity is power. Power is the armor.
The platform doesn’t want to say in advance in which case it bans. Because if it says it, it has to comply. And if it doesn’t comply, a regulator or a court will ask why. Better to say nothing at all. To ban in every case “in our reasonable discretion.” In our sole discretion. No one will ask about the basis.
This isn’t a bug. It’s a legal strategy. It has a name: opacity by design. “Opacity by design.”
Where the construct cracks anyway
This is the weakest of the five layers.
Fair notice isn’t a fiction or an exotic doctrine. It’s a baseline principle of contract law and consumer protection. Two cracks.
First. Consumer regulators. The FTC in the U.S., consumer-protection bodies in U.S. states and in Europe care about deceptive practices. A company says one thing in its public documents and another in its internal operations — that’s a classic deceptive practices case. “We have no prohibition against new accounts” in the public terms + “this violates our policy” in a support email — that is exactly the kind of mismatch the FTC and its peers love to investigate.
Second. The doctrine of “rules promulgated through correspondence.” California law and several other state laws are skeptical of “rules” that appear only in correspondence with the user but impose obligations. In classical contract law this is unilateral modification of contract without notice. Invalid.
What an author should do now
One. Take a “snapshot” of the terms now. Substack’s Terms of Use, Content Guidelines, Publisher Agreement as of today. Save a PDF, a dated screenshot, or an archive.org version. This is your fixation of what was in the terms at the moment of your contract. If the platform later changes the terms and tries to apply the new ones to your past actions — you have evidence.
Two. Save every support letter that names “rules.” If a letter contains phrases like “this violates our rules,” “our policy prohibits,” or “this isn’t allowed” — save it. And check: is that rule actually in the terms? If not, that’s a sign the “rule” was invented at the moment of correspondence.
Three. Don’t explain the terms to the platform on the platform’s behalf. The banned author’s standard reflex is: “look at your own terms, I didn’t violate anything.” Support won’t read that. Regulators in the future will. That’s exactly why it matters to fix the terms as they stood at the moment of your ban.
Four. Keep this in mind: brittle things break under outside pressure. The fifth layer is the most brittle. For a regulator, the absence of a rule in the public documents is a visible, easily verifiable fact. Verified in twenty minutes of reading the terms. That’s good evidence for a complaint.
What this part shows
Five armor layers dismantled. Arc I closes.
§230 immunizes the platform from liability.
Arbitration makes lawsuits impossible.
The First Amendment grants the right to choose.
The closed appeals loop strips complaints of weight.
The silence of the terms lets the platform apply rules that aren’t there.
Each layer is independent. Together they create an architecture in which the author is formally free and practically powerless. Nothing the platform did literally violates explicit rules. It is all “in our reasonable discretion.”
That is the armor. Five layers, five different logics, one system of outcomes.
But at each layer I showed cracks. They’re not random. They’re connected. They add up to a map.
That map is the next arc. Seven parts on the holes in the armor.
— lintara
Next: The algorithm as speech. The 2024 Anderson v. TikTok decision, and why §230 starts to crack precisely where the platform stops being a “pipe” and begins to curate.
Discover more from Lintara
Subscribe to get the latest posts sent to your email.