TikTok, a "cognitive weapon"? The law, the algorithm, and what the evidence says

TikTok was banned in the United States, then saved at the last minute by a forced divestiture. Behind this legal saga, one thesis took hold: TikTok's algorithm is a "cognitive weapon" serving Beijing. The legal side is a dated, documented fact. The cognitive-weapon thesis, however, goes beyond what public evidence establishes. Let us sort it out.

What is established: the law, the Supreme Court, the divestiture

In April 2024, the United States passed the "Protecting Americans from Foreign Adversary Controlled Applications Act" (PAFACA): the law requires ByteDance, TikTok's Chinese parent, to divest the U.S. app or face a ban. On 17 January 2025, the Supreme Court upheld the law unanimously (TikTok Inc. v. Garland), on national-security grounds. The app was briefly pulled, then restored while a deal was negotiated.

On 22 January 2026, the deal closed: TikTok's U.S. operations moved into a joint venture ("TikTok USDS"), majority-owned by a U.S. investor consortium led by Oracle, Silver Lake and MGX; ByteDance keeps only a minority stake (about 20%). The recommendation algorithm is licensed to the new entity, under security oversight. Legal fact: the divestiture did happen, driven by a national-security risk.

Where the concern comes from: ByteDance's legal obligations

The basis for the concern is not a fantasy: it is Chinese law. China's intelligence laws (including the 2017 law) require Chinese companies to cooperate with the state's intelligence work. ByteDance, as a Chinese company, is subject to them. Add to this the opacity of the recommendation algorithm — a "black box" that decides what hundreds of millions of users see. It is this combination, legal capacity plus opacity, that underpins the national-security argument.

The algorithm as a "cognitive weapon": what analysts say

Think tanks (the Jamestown Foundation, among others) analyze the recommendation feed as a potential lever of influence: prioritizing some topics, dampening others, shaping attention at scale. This is the core of the notion of "cognitive warfare": acting on perceptions rather than on bodies. The reasoning is coherent — but it describes a capability and a risk, not a demonstrated operation. Analyzing a lever is not proving it was used.

What this affair proves

• A real, dated U.S. law (PAFACA, April 2024) forced TikTok's divestiture on national-security grounds — a law upheld unanimously by the Supreme Court on 17 January 2025.
• ByteDance is legally subject to Chinese laws that can require it to cooperate with state intelligence work.
• The recommendation algorithm is opaque and does, in fact, shape what hundreds of millions of users see.

What this affair does NOT prove

• That a Chinese influence operation via TikTok has been demonstrated: the law rests on a legally characterized risk, not on a publicly proven case of espionage or manipulation.
• That the app "controls minds": steering attention through a recommendation feed is not decoding or piloting thought.
• That TikTok is a unique case: every large social platform optimizes engagement via opaque algorithms — the difference is the parent company's jurisdiction, not the principle.
• "Beijing can compel ByteDance" (true in law) does not become "Beijing did it via TikTok" (not publicly demonstrated).

So the honest question is not "is China reading your thoughts" — the answer is no. It is "who holds the algorithmic lever of attention, and under which jurisdiction." That is precisely where contemporary cognitive warfare plays out: not in a science-fiction mind control, but in the ownership, opacity and applicable law of the systems that decide what billions of people see every day.

Accueil · Collection · Journal · Sources