On July 6, 2026, at the Bibliothèque nationale de France, Equality Minister Aurore Bergé presented a new national plan against racism, antisemitism, and origin-based discrimination for 2026-2029. Built after six months of consultation with 110 associations, the CNCDH, and the ombudsman's office, the plan runs to 55 measures across four pillars — education, victim support, prevention, and anti-discrimination enforcement. Buried among memorial-site visit targets and police training quotas is a narrower, more consequential line: a commitment to "clarify and harmonize the pathways for reporting online hate, making existing channels — via platforms, PHAROS, Arcom — more legible."
That single measure is worth isolating from the other 54, because it touches the actual machinery of online speech regulation in France, and because France has been here before with worse results.
The Steelman: A Genuinely Confusing System
The case for harmonization is not manufactured. France currently runs at least three separate channels for reporting hateful content: PHAROS, the interior ministry's platform d'harmonisation, d'analyse, de recoupement et d'orientation des signalements, which fielded 109,302 reports in the first half of 2025 alone, with 19,959 of those classified as discrimination; each platform's own in-app reporting tool, required under EU and French law; and Arcom, the audiovisual and digital regulator, which since 2026 has designated nine "trusted flaggers" under the EU's Digital Services Act — including Licra and the Crif, both focused on antisemitism and racism — who get priority handling when they flag content. A user who witnesses a racist post today has no obvious way to know which door to use, whether PHAROS forwards to Arcom, or whether a platform's own report ever reaches either. Consolidating navigation, not authority, is a defensible administrative fix, and Human Rights Watch's July 13, 2026 critique of the plan — which faults it for treating racism as individual prejudice rather than a systemic problem and for skipping disaggregated equality data — does not raise free-expression objections to this particular measure at all, a useful signal that the reporting-harmonization piece is genuinely modest in scope.
The Part the Plan Doesn't Say
But "clarify and harmonize" is doing real work as a euphemism, because France's last attempt to speed up hate-content takedowns didn't survive contact with its own Constitutional Council. The 2020 loi Avia required platforms to remove "manifestly" hateful content within 24 hours of a user flag, and terrorist or child-exploitation content within one hour, backed by criminal penalties. On June 18, 2020, in decision n°2020-801 DC, the Conseil constitutionnel struck down both deadlines, ruling that forcing platforms to adjudicate complex illegality questions under tight, unreviewable timelines — with no judicial check before removal — created "a restriction on freedom of expression and communication that is not adapted, necessary and proportionate to the objective pursued." The core finding: platforms facing criminal exposure for under-removal, but no comparable exposure for over-removal, will systematically over-censor. That is not a hypothetical concern imported from abroad; it's binding French constitutional law, six years old, on this exact policy area.
The 2026 plan's single sentence on online hate gives no indication whether "harmonizing pathways" stays confined to signposting and interoperability between PHAROS and Arcom's dashboards, or becomes a vehicle for the kind of mandatory response-time or removal obligations the Conseil constitutionnel already rejected. DILCRAH's monitoring structure — biannual technical committees, an annual ministerial review, and CNCDH evaluation — governs implementation, but none of the public materials specify legal instruments, meaning the measure could be delivered as a UX and interagency-referral fix, or as the opening line of a redrafted Avia. The distance between those two outcomes is the distance between a plan HRW is right to call modest and a fight the Conseil constitutionnel has already had once.
Why the Distinction Matters Now
The timing sharpens the stakes. Just weeks before the plan's launch, on July 22, 2026, the Court of Justice of the EU issued a platform-liability ruling that EFF warned "could cause collateral damage to freedom of expression" by pushing intermediaries toward more defensive removal. And on August 17, 2026, France's own Constitutional Council blocked a separate bill that would have banned social media for under-15s, again on proportionality grounds — finding the age-verification scheme too broad and insufficiently specified, and sending it back to Prime Minister Sébastien Lecornu's government for a rewrite. That's two proportionality losses for French and EU speech-restricting measures inside a month, both citing the same structural flaw: mechanisms that look procedural but shift enforcement risk onto intermediaries in ways that predictably chill lawful speech.
Against that backdrop, a plan that merely relabels PHAROS and Arcom's existing intake so victims stop guessing which portal to use is genuinely good, low-cost governance — clearer signposting reduces the reporting friction that lets real hate content sit unaddressed while doing nothing to a platform's incentive to over-remove. What France should avoid is using the language of user-experience simplification to quietly reintroduce timeline mandates or liability shifts that its own top court has already found disproportionate twice over. Bergé's ministry has 2029 to fill in the details; the Conseil constitutionnel has already told them, twice, which version they will strike down.