Three weeks before Hungary's April 12, 2026 parliamentary election, the watchdog NewsGuard exposed a covert network of 34 anonymous TikTok accounts churning out AI-generated propaganda to boost Prime Minister Viktor Orbán and smear his challenger, Péter Magyar. The videos ranged from the absurd — talking animals and cartoon toddlers voicing anti-Magyar lines — to the deliberately deceptive: fabricated news broadcasts with AI-generated anchors claiming Magyar had "anger issues," and forged statements attributed to Hollywood celebrities. NewsGuard reported the network drew roughly 10 million views, with 22 of the accounts created in a two-day burst in January 2026. After NewsGuard shared samples, TikTok confirmed the accounts were "part of a covert influence operation that we had previously disrupted" and banned them.
Then Hungarians voted, and Magyar's centre-right, pro-European Tisza party won a landslide — a two-thirds majority of 141 of 199 seats on roughly 53% of the list vote, ending Orbán's 16-year rule (per Wikipedia's tally of the certified results). The disinformation aimed at propping up the incumbent didn't just fail to swing the result; the incumbent it was built to protect lost decisively. That outcome should reshape how the European Union calibrates its disinformation enforcement under the Digital Services Act (DSA).
The strongest case for aggressive enforcement
The case for vigilance is real and should not be dismissed. The same NewsGuard analysis found the Russian influence operation known as Matryoshka seeding false claims — including fabricated reports of Ukrainian cyberattacks on Hungarian infrastructure — across X and Telegram in the same window. The Commission has already opened formal DSA proceedings against TikTok over election-integrity failures tied to the November 2024 Romanian presidential election, where foreign interference was serious enough that Romania's constitutional court annulled the first round. Coordinated, AI-accelerated, often foreign-funded manipulation is a genuine systemic risk, and the DSA was written precisely so that platforms with more than 45 million EU users cannot shrug it off. Nobody serious argues platforms should host covert state-linked deepfake farms unmolested.
But virality is not persuasion
The problem is the implicit theory of harm — that reach equals influence, and that ten million views must translate into shifted votes. Hungary is the cleanest natural experiment we have, and it cuts the other way. As Tech Policy Press argued in its post-election analysis, attributing electoral outcomes to disinformation is "empirically unprovable," and the goal of counter-disinformation work "is not to determine electoral outcomes" but to give voters reliable information. The Russian Storm-1516 content reached at most about 100,000 Facebook users and wasn't even amplified by friendly pro-government media. AI slop was omnipresent on every side, yet organic engagement favoured the challenger. A flood of synthetic video met an electorate that had already made up its mind — and the electorate won.
This matters because enforcement built on a reach-equals-harm premise will systematically over-reach. If the metric of a successful influence operation is impressions, every viral political video becomes a candidate for suppression, and the line between covert manipulation and ordinary, messy, partisan online speech blurs into nothing. The DSA's own March 26, 2024 election guidelines for very large platforms wisely call for measures that are "reasonable, proportionate and effective" while "safeguarding fundamental rights, including the right to freedom of expression." That proportionality language is the whole ballgame, and it should be read as a brake, not an afterthought.
Police the conduct, not the content
The Hungary case points to a sharper enforcement target. What TikTok actually acted on was not the viewpoint of the videos but the conduct: coordinated inauthentic behaviour — 34 sock-puppet accounts, mass-created in days, hiding their operator. That is a defensible, speech-neutral basis for action, and it is exactly where the DSA's systemic-risk framework is strongest. The DSA Observatory's March 2026 assessment found platforms' real failures were not too little censorship but unaddressed structural gaps: mislabelled political ads, undisclosed paid influencers, and opaque algorithmic amplification — problems of transparency and authenticity, not of forbidden ideas.
That distinction should anchor Commission practice. Enforcement that punishes platforms for failing to detect and disrupt covert, inauthentic networks — and for refusing researchers the data to audit them, as the Observatory notes X has done — strengthens the information ecosystem. Enforcement that pressures platforms to remove "disinformation" by content, on the unproven theory that views move votes, invites the predictable failure mode the EFF has documented elsewhere: rushed rules built on shaky evidence that sweep in lawful speech. The Commission should resist the temptation, after a scare like Hungary's, to measure success by takedown volume.
The lesson Brussels should draw
Hungary delivered the rarest thing in this debate: a result. A well-resourced, AI-powered, foreign-flavoured operation ran for months and the side it backed was thrown out by the widest margin since 1989, on record turnout near 79%. The honest reading is that voters are more resilient than the panic assumes, and that the DSA's value lies in transparency, authenticity-verification, and researcher access — not in adjudicating political truth. Police the lie about who is speaking. Leave the contest over what is said to the voters, who, in Hungary, were not fooled.