First Amendment

The First Amendment is an amendment to the Constitution of the United States, adopted in 1791. It prohibits the state from making laws that restrict free speech, the press, the free exercise of religion, or the right to assemble. The text is only one sentence long, yet it is considered one of the most powerful rules in the American legal system. There is an important nuance that is often overlooked: the prohibition is directed against the state, not against private companies. A company may therefore delete posts without violating this constitutional amendment. It is precisely this distinction that makes it so central to the debate about social networks and artificial intelligence.

The reason why US platforms moderate differently than European ones

Almost all major tech companies are headquartered in the United States. Their rules on permitted content therefore arise against the backdrop of American law. There, the latitude for statements is very broad. Even statements that would be criminal offenses in Germany are often protected in the United States.

One example makes the difference clear. In Germany, denying the Holocaust is a criminal offense. In the United States, such a statement usually falls under the protection of the First Amendment, as long as it does not directly threaten anyone. When a US corporation operates a global platform, these two legal views collide.

For the European Union, this has practical consequences. Laws such as the Digital Services Act require platforms to remove illegal content quickly. American politicians and company executives regularly criticize such requirements as censorship. The dispute is therefore not a misunderstanding, but a genuine conflict between two legal traditions.

What the amendment specifically protects – and what it does not

The protection is broad, but not unlimited. Courts in the United States have worked out exceptions over two centuries. Not protected, for example, are direct incitements to violence, genuine threats, fraud, and certain forms of defamation. Everything else is, in case of doubt, considered permitted speech, even if it is offensive or false.

It is also decisive who is acting. If a government agency bans a demonstration, the First Amendment applies. If, on the other hand, a private network deletes a post, it does not apply. One can think of the amendment as a restraint on the state, not as house rules for users on someone else’s servers.

It is disputed whether software outputs also count as protected speech. In 2024, the Supreme Court ruled that a platform’s selection and curation of posts constitutes editorial activity. It therefore enjoys protection. Whether the same logic applies to text generated by a language model is still legally unresolved.

Appearances in news about AI and internet policy

The term comes up as soon as governments propose rules for AI systems. US providers then frequently argue that mandatory requirements for model outputs constitute an interference with protected speech. This argument was also raised regarding planned labeling requirements for AI-generated election advertising.

A second area concerns liability issues. If a chatbot makes a false claim about a real person, that person may sue for defamation. The provider then invokes, among other things, freedom of speech. The first such cases are already underway in the United States and are being closely watched.

For investors and observers of the industry, the connection is concrete. Depending on how courts rule, costs and risks for operators of models and platforms change. Anyone reading reports on internet policy should therefore be able to place this term correctly. It does not describe a general freedom of speech, but a precise American right of defense against state interference.