Ninety-five billion dollars is what the bill signed on Wednesday appropriates, almost all of it for Ukraine, Israel and Taiwan. The provision that may end the American operation of one of the largest consumer applications in the country added nothing to that figure. It costs the government nothing, which is part of why it finally passed.
The Protecting Americans from Foreign Adversary Controlled Applications Act rode into law attached to the foreign aid package, having failed repeatedly as a standalone measure. This is an explanation of what it actually requires, because “ban” is the wrong word for it and the difference decides everything that happens next.
What the law does
It does not prohibit the application. It prohibits app stores and hosting providers in the United States from distributing or supporting it, unless the company executes what the statute calls a qualified divestiture within 270 days. The president may extend that by a further ninety days if a sale is demonstrably in progress.
That construction is deliberate and it is the interesting part. Congress is not telling users they may not use a thing. It is telling two chokepoints, the mobile app stores and the infrastructure providers, that they may not carry it. Regulating a handful of intermediaries is enormously easier than regulating a hundred million people, and it is how most internet-scale rules are actually enforced.
The practical consequence of that design is gradual rather than sudden. An application removed from the stores does not vanish from the phones it is already on. It stops receiving updates, it degrades over months as operating systems move on, and it becomes progressively less usable. There is no day on which the screen goes dark.
What “qualified divestiture” has to mean
The statute requires a sale that ends control by the covered foreign entity, and that is a higher bar than transferring shares.
Control is the operative word and it is the harder one. The test is control rather than ownership, which means the buyer must end up with something that operates independently: no continuing operational relationship of the kind that would let the former parent direct it, no shared data flows that reconstitute the arrangement under another name. Determining whether a proposed transaction clears that is a judgment made by the executive branch, not a box that a filing ticks.
Which raises the question nobody involved has answered in public. What, precisely, is the asset?
The part that is genuinely unresolved
A social application of this kind consists of several separable things: a brand, a user base, a content library, an engineering organization, and the recommendation system that decides what each user sees.
The first four could in principle change hands. The last is the product. The reason people use this application rather than one of its several capable imitators is the accuracy of what it serves them, and that is a property of a system developed over years, by a team, in another country.
So a sale that transfers the brand and the users without the recommendation system sells the least valuable parts. A sale that includes it requires the exporting country to permit the export, and no government is obliged to let a strategic technology leave. Whether Beijing would permit it is the central commercial question in this whole matter and it is not a question American law reaches.
What the enforcement actually looks like
Worth being concrete about who is on the hook, because it is not the company named in the headlines.
The penalties in the statute run against the intermediaries: the entity that makes the application available for download, and the entity that provides hosting for it. Those are large American companies with extensive compliance functions and no appetite whatever for testing the boundaries of a federal prohibition.
Which means enforcement requires no action against users, no blocking at the network level, and no technical measure of any kind. Two or three companies read the statute, take legal advice, and comply. The mechanism is a memo, not a firewall.
Why it passed this time
Versions of this have been attempted since 2020 and have failed, and the mechanics of the failures are instructive.
A standalone bill about one company attracts a coalition of opponents with nothing else in common: people worried about speech, people worried about precedent, people whose constituents use it, and people who simply do not want to vote on it. It also gives every member an easy vote to cast against.
Attached to an emergency appropriation for three wartime allies, it stops being a bill about an application and becomes a clause inside something that has to pass. Members voting for it were not casting a vote about a social network, and members who dislike it could not oppose it without opposing the aid.
That is not a trick. It is one of the few reliable ways a contested measure becomes law in a chamber that cannot pass contested measures, and it is worth naming because it will be used again.
It also explains the shape of the deadline. A statute enforced through two companies needs only a date by which those companies must act, which is far simpler to draft than a rule aimed at a foreign corporation that may not be within reach of an American court at all.
The challenge that is coming
A constitutional challenge is certain and the shape of it is predictable.
The argument will be that this restricts the speech of Americans who use the application to speak, and that singling out one company by name invites the strictest kind of review. The government’s answer will be that this is a regulation of foreign control of a communications platform rather than of content, and that national security supports it.
Neither of those is obviously right. What can be said is that the 270 days will be running throughout, and litigation on this scale does not usually conclude in 270 days.
What is not settled
Nobody has named a buyer. Nobody has stated a price. Nobody has said whether the thing being sold would include the system that makes it worth buying, and the government that would have to approve that has not been asked in public.
What exists is a clock, started on Wednesday, running to a date in January, attached to a transaction that has no parties, no valuation, and no established list of what is on the table. That is an unusual instrument: a deadline for a sale that nobody has yet defined, enforced against two app stores, on behalf of a security concern that the statute itself does not spell out.
