The EU Made Platforms Explain Every Removal

The Digital Services Act usually gets described as a content moderation law. That's accurate, but it misses the most interesting thing it did. The DSA requires platforms to write down why.

Every time an online platform operating in the EU removes a post, restricts an account, demotes something, or disables a listing, it has to file what the regulation calls a statement of reasons. Those statements go into a public database.

That sounds like paperwork. It produced the most detailed public record of content moderation that has ever existed.

What the DSA Covers

A short version first, because the law is broad and the database is one part of it.

The DSA applies to online services used in the EU: marketplaces, social media, app stores, travel and accommodation platforms. It replaced 27 separate national regimes with one set of rules.

Obligations scale with size. Small companies carry lighter requirements. Platforms with more than 45 million monthly users in the EU carry the heaviest, including a duty to assess and reduce systemic risks.

Those risks are listed in the regulation, and the list is worth reading. It covers the spread of illegal content, but also threats to fundamental rights including freedom of expression, threats to media freedom and pluralism, electoral processes, public health, and the protection of minors.

A law that names freedom of expression and media pluralism as risks to be protected, alongside illegal content as a risk to be reduced, is trying to do two things at once. Whether it manages that is a fair question. But the framing matters, because most content regulation names only one side.

The DSA also gives users specific rights: an explanation when content is removed, a route to appeal, a ban on targeted advertising to children, a ban on ads based on sensitive data like religion or sexual orientation, and the option on the largest platforms to turn off algorithmic feeds.

The Statement of Reasons Is the Mechanism

Now the part this article is actually about. Before the DSA, platform transparency meant a quarterly report. A company would publish aggregate numbers: this many removals, this many government requests, this many appeals upheld.

Aggregate numbers are weak evidence. The categories are defined by the company publishing them; they change between reporting periods, and you can't check any individual case against them.

We've written about the limits of voluntary transparency reporting before. The short version is that the record existed because companies chose to produce it, in whatever shape suited them. The statement of reasons works differently. It's filed per decision, in standardized fields, and submission is a legal obligation rather than a goodwill gesture.

Each record includes what was actioned, what the platform did about it, and critically, the grounds claimed.

Three Things the Database Makes Visible

Whether something was illegal or just unwanted

This is the distinction that matters most and the one aggregate reporting hides completely.

A platform can remove content because it breaks the law, or because it breaks the platform's own terms of service. These are entirely different claims. The first says a legal standard was violated. The second says a company decided it didn't want something.

The database separates them. That means you can ask how much of a platform's moderation is law enforcement and how much is editorial preference, and get an answer with numbers behind it.

Whether a government was involved

Records indicate whether a decision came from the platform's own initiative or followed an order or notice from an authority.

That's the thing that was previously impossible to see at scale. A government that sends removal orders quietly, through legal channels, produces no error page and no news story. That kind of censorship-by-paperwork is difficult to measure precisely because it's procedurally normal.

Now there's a record of it, filed by the platform, in a format you can query.

Patterns across platforms and over time

Because the fields are standardized, you can compare. Which platforms remove more under legal grounds than terms grounds. Which categories of content spike in which countries. Whether removals cluster around particular dates.

None of that was possible with quarterly PDFs using different category names.

What It Doesn't Show

I want to be careful not to oversell this, because it has real gaps. It covers EU-facing decisions only. A platform's behavior elsewhere isn't in there.

Data quality depends on platforms filling in fields properly, and the quality varies. Some use free-text fields minimally. Automated submissions can be formulaic in ways that obscure what actually happened.

A per-decision record also isn't self-explanatory. Knowing that something was removed under a legal ground doesn't tell you whether the legal judgment was correct.

And informal pressure stays invisible. A regulator's phone call, a meeting, a quiet warning before a licensing review; none of that generates a statement of reasons. The database captures what was put in writing, which is by definition the part someone was comfortable writing down.

Why This Is Worth Knowing About

I think the DSA database matters less for what it currently reveals than for what it proves.

For fifteen years the argument against mandatory transparency was that it’d be burdensome, technically difficult, and commercially sensitive. The database is evidence that platforms can produce granular, continuous, standardized disclosure when a law requires it. That removes the main objection everywhere else.

There's a fair criticism that the DSA gives regulators substantial influence over speech, and people who find the EU's approach heavy-handed are not being unreasonable. But the transparency requirement is the part of it that cuts in the user's favor, and it's separable from the rest.

A public record of who asked for what to be removed, and on what legal basis, is useful regardless of what you think about the law that produced it. Most countries have nothing like it. They could.