On 19 May 2026, France's audiovisual and digital regulator Arcom unveiled its 2026-2028 strategic project, a twelve-priority roadmap for a "stronger" regulation of the online space. Most of it is unremarkable enforcement housekeeping. Two threads are not. Arcom says it will deploy IP-address blocking — "inédit en France," unprecedented in France — before summer, and that it wants to "simplifier et accélérer les procédures de blocage" of illicit sites, drawing in more intermediaries such as VPNs and alternative DNS resolvers. In parallel, it is widening the perimeter of what it calls "manipulation of information" to cover health misinformation. The direction of travel is toward fast, administrative blocking with the courtroom increasingly out of the loop.
The legitimate core
It is worth stating the regulator's strongest case plainly, because it is real. Coordinated foreign interference, AI-generated synthetic media, and brand-impersonation campaigns now move faster than any court docket. Arcom's own March 2026 assessment of ten major platforms documents how false content is amplified through inauthentic accounts and algorithmic targeting, and roughly 80% of French citizens tell Arcom that fighting misinformation matters. The piracy track record is genuinely good: Arcom credits site-blocking introduced in 2022 with cutting access to pirate services by more than a third. A regulator that can act in hours rather than months is, in narrow and well-defined cases, a feature.
What actually changes: the judge disappears
The problem is the move from court-supervised orders to standing administrative power. Arcom's stated ambition is procedures that are progressively automated and real-time — blocking that can occur, in the regulator's framing, without judicial intervention and without an adversarial hearing. That is a structural shift, not a tweak. It was foreshadowed by the Loi n° 2024-449 du 21 mai 2024 (SREN), which already replaced judicial blocking of non-compliant pornographic sites with administrative orders Arcom can issue directly. The digital-rights group La Quadrature du Net warned at the time that short deadlines plus heavy penalties for non-removal push hosts to "not contest the police and to censor the contents" demanded of them. Automating the pipeline removes the friction that lets that pressure be questioned at all.
France already solved this — with a judge in the room
The striking thing is that France built a proportionate disinformation tool eight years ago and then constitutionally bounded it. The Loi n° 2018-1202 du 22 décembre 2018 on the manipulation of information created an emergency procedure — a référé — but routed it through a civil judge who must rule within 48 hours, and confined it to the three months before a national election. The Conseil constitutionnel upheld the law only after underlining that it could reach allegations of fact likely to distort an election, and explicitly not opinions, parodies, partial inaccuracies, or exaggerations.
That narrow design is not bureaucratic friction; it is the safeguard. A judge ruling in 48 hours is fast enough for an electoral emergency while still forcing someone independent to distinguish a fabricated claim from a contested opinion before content disappears. Arcom's proposed model keeps the speed and discards the judge. It inverts the 2018 settlement: speed was the concession the 2018 law made to the urgency argument, on the condition that an independent magistrate stayed in the loop.
"Health misinformation" is an unbounded category
Expanding the remit to health misinformation compounds the risk. Electoral falsehoods at least have an external referent — a vote happened, a result is verifiable. "Désinformation en matière de santé," which Arcom says is reaching "increasingly concerning" proportions, is a moving target where today's heterodoxy is sometimes tomorrow's revised guidance. Scientific consensus evolves; an administrative body empowered to block "health misinformation" in real time, without an adversarial hearing, is being asked to adjudicate live scientific disputes at the speed of a takedown queue. The category the Conseil constitutionnel was careful to fence off in 2018 — opinion, partial inaccuracy, exaggeration — is exactly the territory health debate lives in.
Overblocking is the predictable failure mode
IP-level blocking is blunter than the URL- or DNS-based measures it extends. Because many sites share an IP address behind a CDN, blocking one can take down unrelated, lawful services as collateral — and pulling VPNs and alternative DNS providers into the enforcement net, as Arcom proposes, widens both the blast radius and the surveillance surface. Automate that and you create a system whose errors are invisible by design: no hearing, no docket, no adversary to flag the over-block before it happens.
A proportionate path exists
None of this requires abandoning enforcement. The proportionate answer is the 2018 model, modernised: keep the speed, keep the judge. A duty magistrate able to authorise emergency blocking within hours preserves the urgency Arcom legitimately invokes while retaining the independent check that made the 2018 law survive constitutional review. Pair that with mandatory transparency — public logs of every blocked IP and the legal basis — and a fast appeal. France does not need to choose between acting quickly and acting lawfully. It chose both in 2018, and the European Convention's free-expression guarantees still apply in 2026. The strategic project's instinct to act is sound; its instinct to act alone is the part that should not survive the summer.