A live-fire year, not a hypothetical one
BENEDMO — the Flemish-Dutch hub of the EU-funded European Digital Media Observatory (EDMO) network, coordinated by the Netherlands Institute for Sound and Vision — published its second annual report on August 5, 2026, covering July 2025 through June 2026. That window happened to contain a snap Dutch general election, called after the coalition collapsed and held on October 29, 2025. So unlike most disinformation-policy documents, this one isn't theorizing about risk; it's grading a real test. Across the year, BENEDMO's fact-check partners — Nieuwscheckers, VRT, Factcheck.Vlaanderen, Knack and Pointer — published more than 500 fact-checks, with clusters around medical misinformation, the Baku climate summit, the NATO summit in The Hague, and, most consequentially, the election itself.
Debunking a party leader's fraud claims in days
The sharpest test came after polls closed. PVV leader Geert Wilders, whose party lost the largest-party position it held after 2023, pushed a viral claim that 15 containers of completed, counted ballots had been loaded onto a trailer in Zaanstad and "disappeared without a trace." BENEDMO's partner network traced the claim to a Facebook account whose owner had already said it was hacked, and the Zaanstad municipal spokesperson confirmed no polling station in the city handles ballot volumes anywhere near that scale. Wilders separately alleged the Electoral Council's vote-counting software had been security-tested by a company founded by a D66 party member — true for the 2023 cycle, but the 2024–2025 testing was actually done by an unrelated firm, Fox-IT. Both claims were checked and published within the same news cycle, not months later in a retrospective report.
The platform-accountability gap the report surfaces
BENEDMO's other election-year finding is less flattering to platforms than to fact-checkers. CampAIgn Tracker, a monitoring tool BENEDMO's network drew on, tracked AI-generated imagery in political posts across Facebook, Instagram, X and TikTok from September 2025 onward and found the PVV was the single largest producer of such content — including fabricated, photorealistic images pairing women with armed, bearded men under captions like "Keep your hands off our daughters!" — with most of it carrying no AI-disclosure label. That's precisely the gap the EU's Code of Practice on Disinformation, folded into the Digital Services Act as a Code of Conduct on February 13, 2025 and binding on signatories from July 1, 2025, is supposed to close through mandatory transparency reporting under DSA Article 35.
Steelmanning the case for binding rules
The strongest argument for that binding approach isn't really about censorship at all — it's about incentives. As Clare Melford of the Global Disinformation Index has put it, "the speech that is actually being suppressed is moderate speech, because it's less profitable," since recommender algorithms reward polarizing content over accurate labeling. If AI-generated fear imagery from a major party spread unlabeled through a national election, as BENEDMO's data suggests, voluntary self-policing by platforms plainly wasn't sufficient. A statutory floor — audited, with real reporting obligations — closes a gap that goodwill alone didn't.
Where the binding-code model still lags
But the comparison this report invites is telling. BENEDMO's decentralized, non-state model caught and killed a false ballot-fraud claim from a sitting party leader within days of an election. The DSA's Code of Conduct, by contrast, only became formally binding in mid-2025, and its signatories' first transparency reports under that new status were published in 2026 — a full audit cycle behind the election it was meant to help protect. The Dutch government's own disinformation policy implicitly concedes the point: its published approach runs on two tracks — citizen empowerment through media literacy, and platform regulation through the DSA — but is explicit that "identifying disinformation is primarily not a task of government," leaving that to independent fact-checkers, precisely the BENEDMO model. Meanwhile, the binding-code route has generated its own friction: U.S. lawmakers led by Rep. Jim Jordan have characterized the Code as a vector for exporting EU speech rules globally, and Brussels had to publicly insist, amid 2026 trade talks, that the DSA was "not on the table" for negotiation — a political cost the fact-checking-network model simply doesn't carry.
The better-tested model deserves the funding
None of this argues against transparency obligations for platforms distributing synthetic political imagery at scale — the CampAIgn Tracker findings show real value in that data existing at all. But BENEDMO's second year is evidence that the fastest, least contested, most electorally relevant disinformation defense in the Netherlands and Flanders this year came from an EU-funded but editorially independent fact-checking consortium, not from a compliance regime still finding its footing. Policymakers weighing where to put marginal funding and political capital — more binding platform mandates, or more of the BENEDMO model — have a year of comparative evidence now. The receipts favor the fact-checkers.